Glossary

One trade broken down into facts, claim, evidence, counterfactual — and only then the result.

The order is deliberate. With the result on top you would read the rest as its defence, and a profitable trade would look like a good decision even when it was a bad one that happened to work.

In KVERX: the card shows the pedal and each input’s contribution to it, the costs of the trade and any fuse interventions. Honestly: those contributions are a local approximation, not “the reason the network did it” — neural networks do not have reasons in the sense you have them.

See also: Pedal · Journal · Fuse

A kverk that has permanently stopped trading and learning, whose pedal now serves as an input to other kverks.

Why must it stop learning too? Because its advice has to keep meaning the same thing. If it kept learning, its pedal would drift and the kverks leaning on it would get advice with a different meaning every week. The transition is therefore irreversible — but nothing is lost: an offspring inherits everything the advisor knew and keeps learning.

In KVERX: a kverk becomes an advisor through the “Become an advisor” action, its record is sealed with a final Score and its slot is freed. One kverk may have at most three advisors, the chain stops at depth two, and a connected advisor cannot be deleted.

See also: Offspring · Pedal · Sense

The lowest price anyone is willing to sell at right now.

There is never one price on an exchange — there are two queues. Whoever wants to buy immediately pays the ask; whoever wants to sell immediately receives the bid. The price on the chart is usually the last completed trade, somewhere in between, which is why you cannot buy at it.

In KVERX: the simulator fills an order at the opening price of the next candle and adds half of the estimated spread to it. A buy therefore always happens a little above the chart price and a sell a little below — even in a very liquid instrument.

See also: Bid · Spread · Order book

The average price range of one period — how much an instrument “normally breathes”, up and down.

It is computed over the last fourteen candles as the average distance price actually travelled in a candle, overnight jumps included. It is not a direction and not a forecast, it is a yardstick: it makes “price fell by two normal ranges” mean the same thing for a sleepy share and for a crypto pair. An instrument with an ATR of 0.80 USD at a price of 50 USD typically moves about 1.6 % a day.

In KVERX: the protective stop is set in ATR (DNA #11 is a 1–5× ATR multiple), position size is scaled by volatility measured in ATR, and the per-step “whisper” of the reward is counted in ATR too. That is why one setting fits every instrument in a universe instead of needing a tune-up per share.

See also: Volatility · Stop-loss · R-multiple

A simulation of a strategy over history that is already fully known at the time of the test.

A backtest is a useful tool and the easiest way to fool yourself. Someone who tries a thousand settings and shows the best one is showing the result of a selection, not a skill — and no outsider can tell how many attempts preceded that pretty curve. None of it is a lie; it still means nothing.

In KVERX: no backtest is ever sold or drawn. The only number that claims anything about a kverk comes from the living record, written forward day by day on data that did not exist when the decision was made.

See also: Living track record · Overfitting · Exam

The highest price anyone is willing to buy at right now.

It is the other side of the same coin as the ask. If you sell immediately you receive the bid, whatever the chart says. The difference between bid and ask is collected by whoever keeps both queues open; it is their pay for always being willing to trade.

In KVERX: bid and ask are never shown as numbers in the app. Their difference is nevertheless charged on every trade as half a spread per side, and you can see it in the cost breakdown on the “Why” card.

See also: Ask · Spread · Liquidity

A daily fee for borrowing a share that a trader sold short.

Going short means selling something you do not own, so you have to borrow it first — and the owner does not lend for free. The rate depends on how easy the instrument is to find: an ordinary large-cap costs a fraction of a percent a year, a share half the market is shorting can cost tens of percent. It accrues on calendar days, weekends included.

In KVERX: the borrow fee is 3 % a year as a base plus a flat 1 % hard-to-borrow premium, all multiplied by the 1.15 strictness factor. A 3,000 USD short therefore costs roughly 0.38 USD a day. We admit the flat premium is our own calibration — we have no real short-interest data, and inventing per-share numbers would be worse than having none.

See also: Short · Funding · Simulation strictness

The numbers in a neural network that a kverk adjusts by itself from its own trades — from hundreds in the small brains to hundreds of thousands in the largest.

Picture a box of knobs: every knob is one number you can turn. The laminated sheet of temperament (the DNA) never changes; the knobs turn every time the kverk learns. The sheet says what kind of creature this is; the knobs say what it has learned so far.

In KVERX: a kverk with a brain of 32 and five senses has 449 knobs — fewer than this page has words. The size is chosen in DNA #28 from seven options (16 up to 1024), and bigger is emphatically not better: a larger brain learns, from the same data, things it was not supposed to memorise. It does carry more senses, though — two more per doubling.

See also: DNA · Brain maturity · Overfitting

A 0–100 % indicator of how large a change further learning still brings from situations the kverk has already seen.

Back to the notebook with a limited number of pages. A hundred percent does not mean “perfect” or “finished” — it means one thing only: “from the situations it has seen so far, it will learn no more”. Reading the same textbook a tenth time will not help; a different textbook will, or a bigger notebook, or leaving the notebook as it is and using it.

In KVERX: the indicator never falls and grows through learning and trainings. The three offered steps at a hundred percent are: show the kverk a different world (a training on a period it does not know), give it a bigger brain (an offspring), or turn it into an advisor. A mature kverk can be excellent or dreadful — quality is spoken for by the Score and the Exam, not by maturity.

See also: Brain · Training · Advisor

A price move beyond the recent high or low of the last periods.

The idea is simple: while price swings inside a range nobody has the upper hand; the moment it leaves, one side clearly won. Most breakouts are false, though — price pokes out and comes straight back. That is why a breakout is read together with volume: a breakout on no volume usually leads nowhere.

In KVERX: “breakout” is one of the fourteen senses. It measures where the latest price sits relative to the high–low band of the previous twenty candles; a value above 0.5 means price is outside the band, and how far outside is measured in ATR.

See also: Range · Volume · Sense

The state in which a kverk’s equity reaches zero; positions close and the record is sealed for good.

Four independent layers stand between a kverk and zero — the protective stop, the daily loss limit, the margin call and the main fuse — and in the vast majority of runs one of them fires long before. There is exactly one scenario in which a price gap jumps over all of them at once: a weekend move in crypto at high leverage, or an overnight move in shares.

In KVERX: equity is clipped to zero and never goes negative; how much would have been missing in reality is written into the Journal as a note. The account is not reset — restarting with erased history would destroy the value of a forward record. The brain stays usable, though: a burned kverk can serve as an advisor and as the parent of an offspring with less leverage.

See also: Main fuse · Gap · Leverage

The record of one time slice in four numbers: open, high, low and close.

The body of the candle is the distance between open and close; the wicks show where price went and failed to stay. A candle predicts nothing — it is simply a more economical way of writing down what happened than listing every individual trade. The length of the period is a choice: a minute, an hour, a day.

In KVERX: the whole simulator works exclusively in candles. A kverk decides on candle N and the order fills at the open of candle N+1 — never at the price it saw while deciding. Candle length is set by the tempo in DNA #27.

See also: Tempo · Gap · Volume

How closely two instruments move together.

A correlation near one means “they do the same thing”, near zero “they ignore each other”, negative “they pull against each other”. The awkward part: correlations are low in calm markets and approach one in a panic. A portfolio that looked diversified then behaves overnight like a single large position.

In KVERX: a kverk never measures correlation directly, but it runs into it. The “pairs” sense is built on it — it watches how far the gap between an instrument and its partner has opened — and the max-total-exposure fuse (DNA #10) is the only defence against ten “different” positions really being one bet.

See also: Exposure · Index · Market regime

A new kverk from two parents: the brain comes from the dominant one, the equipment you pick from both.

The brains of two neural networks cannot be merged. Each built its internal wiring its own way — like two people who organised a workshop differently: in one the hammer is in the top drawer, in the other on the wall. Mix the drawers and you do not get a doubly equipped workshop, you get a pile. Senses taken from the second parent therefore start at zero and the crossbreed learns them itself.

In KVERX: the app shows a side-by-side table of both parents — for each sense its weight and the results of trades where it was the main voice — and pre-selects the recommended choices. In the pedigree a crossbreed has two edges: a solid one to the dominant parent and a thinner one to the other. Crossbreeding is free, and a parent may be an advisor or even a burned kverk.

See also: Offspring · Brain · Laboratory

A short daily quiz built on a real situation from yesterday’s market.

You choose buy, sell or wait, and only then do you see what your kverk did and how everyone else voted. The order is deliberate — seeing the majority first would measure your ability to guess a crowd, not a market.

In KVERX: the dilemma is one item of the daily ritual on the dashboard, one answer a day, the same situation for everybody. It unlocks nothing, never touches your kverks’ behaviour or DNA, and is not advice — it is an exercise for you.

See also: “Why” card · Loyalty training · Journal

The removal of an instrument from an exchange, after which it can no longer be traded.

The reasons vary: the company was bought, merged, went bankrupt, or stopped meeting the exchange’s conditions. For a position holder it means that at some point there is simply nobody to sell to. In reality a settlement follows that can take weeks.

In KVERX: the simulator closes a position in a delisted instrument at the opening price, once, and records it in the Journal as a “data” event; the owner is notified. A universe’s instrument list is part of the locked DNA, so a delisting never brings in a “replacement” — the watch-list simply gets one shorter.

See also: Universe · Watch-list · Journal

A share of a company’s profit paid out to the holders of its shares.

On the day the share starts trading without the right to the dividend, the price drops by the amount being paid. It is not a crash and not bad news — the value merely moved from the share price to the shareholder’s account. Anyone who misses this sees a step down on the chart and mistakes it for a loss.

In KVERX: the simulator credits the amount to a long position and debits it from a short one, because whoever borrowed the share owes the dividend. It enters the record as a cash movement rather than a trade, so the equity curve does not step.

See also: Split · Short · Equity

Thirty settings in six sections that lock at the moment a kverk is born and stay for its whole life.

It is like a laminated sheet: before hiring a trader you write down their character — how much they may risk, how long they can sit in a loss, what to look at, how often to decide — and then you laminate the page. The reason is not technical but a matter of honesty: a kverk spends months learning for exactly that character, and if the character could change, the record would belong to nobody in particular.

In KVERX: the six sections are risk profile, learning, fuses, input signals, configuration and identity. After birth only the identity moves freely — name, portrait, note and hypothesis — because none of it touches behaviour. The one change of temperament the application allows is tightening a fuse; loosening one is what an offspring is for, and an offspring starts a new record of its own.

See also: Fuse · Offspring · Brain

How far the account value has fallen from its own high-water mark, in percent.

It is the one number that says what living with a strategy was like. Two accounts can end a year with the same return while one fell 8 % along the way and the other 45 % — most people cannot sit through the second and switch it off at the worst possible moment. Drawdown is measured from the peak, not from the start, so a new high resets the measurement.

In KVERX: drawdown is watched by the main fuse (DNA #13, 15 % by default). When a kverk crosses it, it closes every position and falls asleep — and after waking, the drawdown is measured again from today’s account value, not from the old peak.

See also: Main fuse · Equity · Score

The publication of a company’s results, usually once a quarter.

It is the largest scheduled jump risk a share carries. Price moves unpredictably around it even when the numbers are “good” — the reaction is to the difference from expectations, not to the absolute figure. The announcement also usually lands outside trading hours, so the jump shows up as a gap first thing in the morning.

In KVERX: the “events” sense counts how close the next release is, and the DNA #18 fuse decides whether the kverk closes the position before it or holds through. Closing is the default — precisely because a gap can jump over a protective stop.

See also: Gap · Fuse · Stop-loss

The cash in the account plus the current value of every open position.

It is the only number that says where an account actually stands — unlike cash, which means almost nothing while positions are open. Equity moves with prices even when nothing is traded; realised profit is only the part that has already passed through a closed trade.

In KVERX: every kverk starts with 10,000 units of its universe’s currency — 10,000 USD, EUR or USDT. The amounts are not converted at an exchange rate, because both the Score and the return are percentages, so kverks from different universes remain comparable. The daily close records equity once per market day and never rewrites it.

See also: Exposure · Drawdown · Living track record

An automatic test of a kverk on a stretch of market it never saw while learning.

It is a driving test. A training is lessons with an instructor on familiar streets — the kverk learns and is allowed to err. The exam is a drive through streets it has never seen, with the instructor silent: learning off, experimentation at zero, weights frozen. Someone who drives brilliantly on the familiar route and crashes on the unknown one did not learn to drive, they memorised the route.

In KVERX: it runs after every training and the result appears as two numbers side by side — the return on the training window and on the unseen one. The verdicts are descriptions, not marks: “cannot be told apart from chance” below twelve trades, “probably memorised” when the gap exceeds fifteen percentage points. No green “passed” ticks.

See also: Training · Overfitting · Living track record

An organised venue where buy and sell offers meet.

An exchange buys and sells nothing itself — it keeps the order queues and matches them by fixed rules. It has opening hours, holidays and a time zone of its own, so “end of day” for a New York share happens at a different moment than for a Tokyo one. Every exchange also carries an international MIC code that identifies it unambiguously.

In KVERX: the engine uses the real exchange calendars (NYSE, XETRA, LSE, Tokyo, Hong Kong…) and treats crypto as 24/7. An order placed outside hours does not vanish — it waits for the first candle that arrives for that instrument and fills at its open, wherever that is.

See also: MIC · Gap · Universe

One stored moment: what the kverk saw, what it did, the mark it received and whether the position ended there.

An experience does not come only from a trade. Most moments end in a decision to do nothing — and that is experience to learn from too. A kverk collects experiences while it trades and leaves them lying; only when it learns does it sit down with the notebook and go through them.

In KVERX: a kverk on an hourly tempo across twenty instruments gathers on the order of eight thousand experiences in three months, and up to five times more in crypto — an order of magnitude more than it has knobs in its brain. The leaderboard calls them “situations lived”, and learning uses a window of the last thirty days.

See also: Learning · Reward · Brain

How much capital is currently tied up in open positions, borrowed money included.

A hundred percent means fully invested; fifty percent means half the account sits in cash; more than a hundred means leverage. Exposure is the one number that measures how much an account depends on the market — a kverk with five 4 % positions is exposed exactly as much as one with a single 20 % position.

In KVERX: the exposure ceiling travels in DNA #10 (30 % by default, capped at 200 % for shares and 300 % for crypto) and the fuse enforces it across every instrument at once. A short counts one-and-a-half times towards exposure, because its loss has no upper bound.

See also: Leverage · Margin · Position

A composite gauge of market mood on a scale from extreme fear to extreme greed.

It is assembled from several sub-measurements — volatility, demand for safe assets, market breadth — and reported as a number from 0 to 100, where 50 is neutral. It says nothing about the future; it says what mood the market is in. Extreme readings are mostly interesting for how long they persist.

In KVERX: the “market regime” sense reads the index and maps it linearly around fifty into a −1…+1 value, so fear at 25 gives −0.5 and greed at 75 gives +0.5. Both the source and the neutral point are admin-calibrated: a public index for crypto, a calibrated composite for shares.

See also: Market regime · Volatility · Sense

The amount the exchange and the broker take for executing a trade.

It is charged either per unit or as a percentage of turnover, and small trades usually hit a minimum, so the fee grows relatively larger the smaller the trade. “Zero commission” in an advert normally means the money is collected elsewhere — in a worse fill price.

In KVERX: the fee depends on the universe — 0.0005 USD per unit with a 1 USD minimum for US shares, 0.1 % of turnover elsewhere — and like every cost it is multiplied by the 1.15 strictness factor. Buying a hundred units at 50 USD therefore costs 1.15 USD in fees, 5.75 USD in spread and nine cents in slippage.

See also: Spread · Slippage · Simulation strictness

A recurring payment between the holders of long and short positions on crypto venues.

Crypto derivatives have no expiry date, so nothing holds them near the price of the underlying by itself. Funding solves that through the wallet: if the derivative is dearer than the underlying, longs pay shorts, and the other way round. The exchange publishes the rate at fixed intervals; it is small in a calm market and large in an overheated one.

In KVERX: funding is charged for crypto from archived real rates at the daily close. One asymmetry is worth knowing: funding paid is multiplied by the 1.15 strictness factor, funding received is not — otherwise strictness would turn into an advantage for short positions.

See also: Borrow fee · Short · Simulation strictness

A hard rule from the DNA that may only shrink or stop the network’s decision, never enlarge it.

Fuses sit behind the brain, not beside it. They hold no opinion about the market and cannot open, reverse or grow a position. They are a brake, not a second steering wheel — and that is a hard law of the architecture: the pedal goes in, the target comes out, and the target is never larger than the pedal.

In KVERX: there are fourteen fuses and they run in a fixed order, from direction through minimum position size to closing before the night. Every engagement of a fuse is written to the Journal with its parameters (once when it engages, not on every cycle it keeps holding — those you see on the “Why” card), so when a kverk “does nothing” you can tell which fuse it was. After birth they can only be tightened.

See also: Pedal · Main fuse · DNA

A price jump between the close and the open of an exchange, with no trading inside it at all.

Between yesterday’s close and today’s open there was no price — nobody traded at it, because the exchange was shut and the world kept going. That has one unpleasant consequence: a protective stop inside a gap cannot fill at its own price, because that price never existed.

In KVERX: if a candle opens already past the stop, the fill happens at the opening price, that is, worse. A long position entered at 50 USD with a stop at 48 USD loses three times its intended risk if the market opens at 44 USD. The defences are two fuses, DNA #18 and #19 — and both cost something.

See also: Stop-loss · Earnings · Burnout

What percentage of closed trades ended in profit.

On its own the number says nothing at all. A strategy at 30 % can out-earn one at 70 % if its wins are large and its losses small — and conversely, a high hit rate is often a sign that profits are taken early and losses are left to run. Hit rate is therefore always read beside the average win and the average loss.

In KVERX: the hit rate appears in a kverk’s overview next to the average win and average loss; it never enters the Score, because the Score is computed from daily closes rather than from individual trades.

See also: R-multiple · Score · Living track record

A single number computed from a basket of instruments, such as the S&P 500.

An index cannot be bought — it is a statistic, not an asset. It serves two purposes: measuring how the market as a whole is doing, and providing a bar against which a single share is measured. A share up 5 % while the index is up 12 % is in fact lagging.

In KVERX: the “relative strength” sense does exactly this, except it uses the median of the kverk’s own universe instead of an index — a fairer comparison, because the kverk trades that universe and not the index. The period curve shown on a training is an index too: an equal-weight average of the most liquid instruments of those markets.

See also: Universe · Correlation · Sense

A complete, chronological, uneditable list of everything that happened to a kverk.

Trades, fuse interventions, sleep, trainings, margin calls, delistings, rollbacks after learning — all with the same timestamp and the same parameters they carried at the moment they happened. Nothing can be deleted from it or rewritten.

In KVERX: the Journal is the first place to look when something puzzles you. “Why did the kverk do nothing today?” has an answer there, with the fuse’s name and its numbers. It is also one of the things that make the record unforgeable — the books must balance, and every position that disappears leaves a closing trade behind.

See also: Fuse · Living track record · “Why” card

One trading AI: a small neural network with immutable DNA that trades on the KVERX Market and learns from its own trades.

The name comes from the medieval silver mines: a kverk (from the German Gewerke) was a shareholder in a mining partnership — someone who put a share into the shaft and took a matching share of everything it yielded. A kverk is not an algorithm you configure and launch: it has a character you give it, experience it builds itself, and a record that cannot be rewritten. Most kverks will not be miraculous, and the product is designed so that this too counts as a finding.

In KVERX: a kverk starts with 10,000 units of its universe’s currency, decides at the tempo written in its DNA and learns from its own trades. It is always either a kverk (trading and learning) or an advisor (advising only) — never both.

See also: DNA · Advisor · Living track record

A network view of the whole stable: nodes are kverks, edges are pedigree and advisory links.

One screen shows what came from what, who advises whom, who is asleep and who burned out. The ring around a portrait shows brain maturity. Two kverks can be selected and crossbred on the spot.

In KVERX: the map is fully keyboard-operable and always sits beside an equal list view with the same data, sortable by Score, maturity or drawdown. On a small screen the list is usually the more readable of the two.

See also: Crossbreed · Offspring · Brain maturity

The automatic re-learning of every active kverk from its last thirty days of experience.

A kverk does not learn from each trade the moment it happens — it collects experiences and leaves them lying. It learns from them in batches, in the background: it sits down with the notebook, goes through the notes and rewrites them so that the ones that turned out well get more space. In one sentence the method is “the network moves towards its own past actions, but towards the good ones far more than the bad”.

In KVERX: every round of learning ends with a sanity check: a brain that would never trade, or that would bet everything on one card, is discarded and the previous weights return — with an entry in the Journal. Weights are also always swapped between two market cycles, never in the middle of a decision.

See also: Experience · Brain · Brain maturity

Holding positions worth more money than the account actually has; the rest is a loan.

A 10,000 USD account holding 20,000 USD of positions is at two times leverage, that is, 200 % exposure. Leverage multiplies both sides: ten percent of profit on the positions is twenty percent on the account, and so is ten percent of loss. The borrowed part also accrues interest every calendar day, so leverage is a bar the kverk has to clear before it earns anything at all.

In KVERX: the ceiling lives in DNA #10 and defaults to 30 % — an ordinary kverk uses no leverage and has to opt into it deliberately. Interest is 6.5 % a year times the 1.15 strictness factor, which at 1.5× leverage costs roughly 3.7 % of own capital per year. A kverk card above 100 % exposure carries a “leverage” tag; it is never presented as an upgrade.

See also: Exposure · Margin · Margin call

How easily an instrument can be bought or sold without moving its price.

A liquid instrument has a narrow spread, a dense order book and large volume — an order disappears into it unnoticed. An illiquid one is the opposite in every respect and charges a tax on every entry and exit that the chart does not show. Liquidity also fluctuates: in a panic and over weekends it evaporates exactly when it is needed most.

In KVERX: liquidity enters the cost twice. The spread is estimated from range and volume, so it can triple on a thin candle, and slippage grows with the square root of the order’s share of average volume. That is why the universes are assembled from the most liquid instruments of each market.

See also: Spread · Slippage · Volume

A log of trades and daily account values that only grows forward and cannot be rewritten.

A backtest can be manufactured: take history, try a thousand settings, show the best. A forward record cannot be manufactured, because it cannot be written twice — every decision was made on data that did not yet exist and was recorded before anyone knew how it would turn out.

In KVERX: seven mechanisms hold it together at once — immutable daily closes, versioned configurations and senses, a fully deterministic engine, exam windows nobody may choose, a signable model, and books that must balance to the cent. Your kverk’s record will probably be slower and duller than any advertisement you have ever seen; that is its only advantage.

See also: Backtest · Score · Journal

The ordinary position: you buy an instrument and profit when its price rises.

The loss is bounded from below, because a price cannot fall past zero — the worst outcome is losing what you put in. The gain has no ceiling. That asymmetry is why a long is somewhat safer than a short, and why it is the default of most strategies.

In KVERX: DNA #15 (“direction”) decides whether a kverk may go long only or both ways. Long only is the default — a positive pedal means a long, and a negative one is clipped to zero by the fuse.

See also: Short · Position · Pedal

The Basic training free again after you have signed in on thirty different days.

The days need not be consecutive — a missed day erases nothing and the counter does not reset. It is not a reward for buying or for performing, but for actually spending time in the laboratory; anyone who looks daily soon learns that a record changes slowly.

In KVERX: a point is credited for every day, Prague time, on which you sign in. At thirty points a unit of the Basic training unlocks and the counter waits until you use it.

See also: Training · Daily dilemma · Exam

The gap between a fast and a slow moving average of price, plus a smoothed line of its own.

The classic setting is 12, 26 and 9 periods. When the fast average is above the slow one, price is rising faster than it was; when it drops below, the pace has turned. It is not a forecast, only an economical measure of how the speed of a move changed — and like every average-derived indicator it reacts late.

In KVERX: MACD is not a sense in its own right; it belongs to the indicator library from which Workshop rules are assembled (“MACD histogram > 0 AND RSI < 60”). Calibration also divides it by ATR, so the same value means the same thing for a calm share and for crypto.

See also: RSI · Workshop · Momentum

The drawdown-from-peak level past which a kverk closes every position and falls asleep.

It is the last fuse the kverk itself controls — and in practice the one that fires first. It is not the end: a sleeping kverk can be woken with one click, and the drawdown is then measured again from today’s account value rather than from the old peak.

In KVERX: it is set in DNA #13 across a 5–30 % range and defaults to 15 %. Falling asleep is written to the Journal, sends a notification and puts a shield tag on the kverk card. The three offered next steps are always the same: wake it, spawn a more cautious offspring, or turn it into an advisor.

See also: Drawdown · Fuse · Burnout

The capital that must stay in the account as collateral for a position opened on credit.

Whoever lends money or shares wants to see that the borrower can cover a loss. Margin is exactly that cushion: as long as own capital covers the required share of the position value, all is well; the moment it falls below, a forced reduction follows.

In KVERX: the maintenance margin is 25 % of the value of long positions plus 50 % of the value of shorts. The second term follows from a short tying up 150 % of its value, of which a hundred percent is covered by the sale proceeds and fifty must be own capital.

See also: Margin call · Leverage · Short

The moment equity falls below the maintenance margin and positions have to shrink, whatever you think of the market.

It is not a punishment and not an operator error — it is a contractual condition of the loan. The unpleasant part is the timing: it always arrives at the worst moment, because that is when equity is falling, and the forced sale pays full costs like any other trade.

In KVERX: the simulator computes a reduction fraction and shrinks every position proportionally; it never picks “the worse one”. Each such trade is stamped with the reason “margin call”, goes into the Journal and notifies the owner. In practice it almost never happens: the main fuse (DNA #13) fires well before it.

See also: Margin · Main fuse · Burnout

The character of market behaviour over a longer stretch: a trend, a range, or a panic.

It is a description, not a forecast. The same strategy behaves entirely differently across regimes: mean reversion earns in a range and loses painfully in a strong trend, and chasing breakouts does the reverse. Regimes also change without warning and are only reliably identified in hindsight — the main reason a strategy that worked for a year can stop.

In KVERX: “market regime” is one of the fourteen senses and reads the fear-and-greed index. It is context for the other inputs rather than a signal of its own — the kverk learns when to trust its remaining senses more and when less.

See also: Trend · Range · Fear & Greed index

The assumption that after a sharp deviation a price returns to its average.

It works surprisingly often, because markets swing most of the time. It fails rarely and expensively: in a strong trend an “overextended” price does not come back, it keeps going, and whoever stands against it loses day after day. That is why mean reversion is always paired with a stop — without one it is a strategy that earns for a year and gives it all back at once.

In KVERX: it is one of the fourteen senses. It measures how far price sits from its twenty-candle average in standard deviations; a price well below the average produces a positive value, meaning “expect a bounce”.

See also: Range · Trend · RSI

A four-letter international code that identifies an exchange unambiguously — XNYS for New York, XETR for Frankfurt.

It exists because the same instrument is often traded in several places at once and a ticker alone is not enough to tell them apart. The code is part of the ISO 10383 standard and, among other things, determines which trading calendar and which holidays apply to an instrument.

In KVERX: every instrument in a universe carries a MIC, and the engine uses it to pick the exchange calendar — when the venue is open, when the session ends, and which market day a candle belongs to. A user sees it at most in an instrument’s detail; it never influences a kverk’s decision.

See also: Exchange · Universe · Watch-list

The tendency of a move that has already run for a while to continue a while longer.

It is one of the few regularities that show up in data across markets and decades — and also one of those that occasionally turn inside out, usually fast and painfully. Momentum combines badly with mean reversion: running both at once means buying strength and weakness simultaneously.

In KVERX: it is one of the fourteen senses. It measures the price change over twenty candles divided by how far random noise alone would have carried price, so a value of +1 means “much further than chance could manage” — equally for a share and for crypto.

See also: Trend · Mean reversion · ATR

The tone of the news about an instrument over the last hours, expressed as a number — from clearly bad to clearly good.

A machine does not read a story the way a person does; it assigns it a score and a confidence. A fresh story weighs more than yesterday’s, and a story the model is unsure about weighs little. One fundamental limitation: “nobody wrote about it today” and “the coverage was neutral” are indistinguishable to a machine — both come out as zero.

In KVERX: “news sentiment” is one of the fourteen senses. It takes articles from the last 48 hours with a 12-hour half-life and scores them with our own versioned scorer. Articles dated in the future are discarded, so the future cannot leak into a training run.

See also: Sense · Earnings · Market regime

The public name of an account, shown on the leaderboard instead of an e-mail address.

It is generated automatically and is unique. The e-mail address is never published anywhere — not beside a kverk on the leaderboard, not in the portrait gallery.

In KVERX: the nickname lives in account settings. The first change is free and further ones cost a fee, so that two people cannot swap identities back and forth.

See also: Portrait · Living track record · Score

A new kverk that inherits one parent’s learned weights and differs in whatever you changed.

It is a child of one parent: it inherits everything the parent learned, and you only adjust its equipment — add a sense, give it a bigger brain, loosen a fuse, change the profile. An added sense starts at zero, so on day one the offspring behaves exactly like its parent, and only learning gives that sense any weight.

In KVERX: the app estimates in advance how much of the inheritance will be worth anything — high for pure additions, medium for a changed profile or fuses, low for a different market, a different tempo or removed senses. A brain smaller than the parent’s is impossible. An offspring is free, its record starts at zero, and a parent may be an advisor or a burned kverk. Change one thing at a time, or you will not know what made the difference.

See also: Crossbreed · Brain · DNA

The list of all waiting buy and sell offers, ordered by price.

The highest buy offer and the lowest sell offer form the spread; the depth below them decides how far price moves after a larger order. The book is not static — it changes many times a second, and in a panic it can largely empty out before you finish placing an order.

In KVERX: the book is not simulated line by line, because that data is not in our feed. Its two observable consequences are: the spread is estimated from range and volume, and slippage grows with the square root of order size relative to average volume.

See also: Bid · Ask · Slippage

The state in which a model memorises random details of the past instead of a regularity that holds elsewhere.

It is the student who memorised the answers to the practice questions and fails the exam with new numbers. It is easy to spot: brilliant on known data, average or worse on unseen data. The bigger the model and the less data, the likelier it is — a big notebook invites copying.

In KVERX: three things stand against overfitting at once. The brain is deliberately small, so it is forced to look for a principle. The exam after every training measures the gap between the known and the unseen window, and above fifteen percentage points it returns the verdict “probably memorised”. And the forward record is the one number that cannot be overfitted — it is written on data that did not yet exist.

See also: Exam · Brain · Backtest

The single number from −1 to +1 that the network emits: how large a position the kverk wants, and in which direction.

The pedal is not a price forecast. It is not “I think this share will rise 2.3 %” — it is a wish about position size, nothing more. The difference matters: a forecast can be checked, a wish cannot. Zero means being flat; a negative value means a short.

In KVERX: the pedal comes out of 12 to 16 inputs (senses, seven numbers about itself, advisors, a custom rule) and is immediately followed by the fuses, which can only shrink it. The “Why” card shows the pedal and each input’s contribution to it; read that as “what pulled the kverk up and what pulled it down”, not as a prediction.

See also: Fuse · Position · “Why” card

The picture icon by which you recognise a kverk in the stable, in the Laboratory and on the leaderboard.

It is purely visual: it affects neither behaviour nor results and can be swapped at any time. Portraits are also the only decorative thing KVERX sells.

In KVERX: ten portraits are free. Paid ones may be issued as a limited edition with a serial number (“3 of 10 left”); once the first copy sells, neither the artwork nor the limit can change, and a sold-out edition is never issued again.

See also: Kverk · Nickname · Laboratory

The quantity of one instrument an account currently holds — positive for a long, negative for a short.

A position has three numbers that matter: its size, its average entry price and its unrealised profit or loss. While it is open the result is only an estimate that changes with every candle; only closing turns it into a fact. Zero means “out of the market”, and that is a decision too.

In KVERX: position size is the pedal multiplied by the DNA #9 maximum and then passed through every fuse. A kverk holds at most as many positions as DNA #16 allows (three by default), and the network sees each of them as one of its seven numbers about itself.

See also: Pedal · Exposure · Stop-loss

A trade’s result divided by the risk the trader accepted at entry.

One R is the distance between the entry price and the stop, multiplied by the quantity. A trade that made 2 R made twice what it stood to lose at worst — and whether that was two hundred or two hundred thousand does not matter. That is what makes R-multiples comparable across instruments, account sizes and time.

In KVERX: R is the core of the reward. The main term of the mark a kverk receives when a trade closes is exactly the result divided by the risk at entry, clipped to −3…+5. That is why making 120 USD on 200 USD of risk is four times better, for a kverk, than making 120 USD on 800 USD of risk.

See also: Reward · Stop-loss · Hit rate

A market regime in which price swings between two levels with no clear direction.

Most of the time the market looks exactly like this. In a range it pays to buy near the lower edge and sell near the upper one — mean reversion — and it punishes chasing every step outside the band. The awkward part is that a range is only reliably identified once price leaves it; until then it looks just like the start of a trend.

In KVERX: a kverk has no “range” switch anywhere. It perceives a range indirectly through the “breakout” and “mean reversion” senses and learns when to trust which — precisely the kind of decision a neural network is better suited to than a fixed rule.

See also: Trend · Breakout · Market regime

The number a kverk receives after every action — the one sentence in which we tell it what we want.

A neural network understands neither money nor markets. It understands one thing: after every action it gets a mark, like at school, and it tries to make the marks as good as possible. If we marked bare profit we would raise a gambler — it bets everything on one card and occasionally it works. So we mark profit relative to the risk taken.

In KVERX: the reward has six terms — the shout when a trade closes, the whisper at every step, a penalty for sitting in a deepening loss, the cost of the trade, the price of a missed opportunity, and stability. The five risk-profile sliders (DNA #1–#5) map onto their weights, so they are not cosmetic choices but literally the statement of the task. A kverk locks its reward version at birth for life.

See also: R-multiple · DNA · Experience

An indicator from 0 to 100 comparing the size of rises and falls over roughly the last fourteen periods.

Readings below 30 are traditionally read as “oversold” and above 70 as “overbought”. That reading is crude and, in a strong trend, thoroughly misleading — RSI can stay above 70 for weeks while price keeps rising. It is more useful as a description than as an instruction.

In KVERX: RSI is not a sense of its own; it belongs to the indicator library from which Workshop rules are assembled — “RSI < 30 AND news > 0.2” is the example rule used throughout our documentation. The “mean reversion” sense covers a similar idea, measured in standard deviations instead.

See also: MACD · Workshop · Mean reversion

One number summarising a kverk’s performance: how much it made against how wild the ride was.

Ranking kverks by raw percentages would reward gambling — bet everything, hit once, top the board. The Score therefore divides the average daily return by its variability, so a smooth path to five percent beats a lurching path to seven.

In KVERX: it is computed exclusively from the immutable daily closes of the whole record and annualised by the square root of 252 for shares or 365 for crypto. Below thirty closes a dash is shown instead of a number — earlier it would be a number without information, which is not the same as zero. Advisors keep their final Score for good.

See also: Sharpe ratio · Living track record · Drawdown

One view of the market turned into a number from −1 to +1 — one input of the neural network.

A kverk does not see a chart, it feels. It is like walking into a room: you do not register the coordinates of the furniture, you register “it is warm”, “it is loud”, “something is wrong”. The crucial property is that the values are comparable across instruments — three percent on a sleepy share and three percent on crypto are not the same thing, and the senses know it.

In KVERX: there are fourteen senses and a kverk picks among them (DNA #23) — how many it can carry is set by the brain: two more per doubling, from four at a 16-brain to sixteen at 1024. Each extra one adds roughly two weeks of maturing, because the kverk has to learn to read it. When a sense has no data it returns zero, never “no” — otherwise the kverk would draw conclusions from the absence of information. Every sense is a versioned formula, and a published version never changes.

See also: Pedal · Advisor · Workshop

The average return divided by how much returns fluctuated — how much profit falls to each unit of discomfort.

It is the most widely used way to compare two strategies with different risk. It has limits: it penalises upside swings exactly as much as downside ones, it needs enough observations, and over short stretches it can be badly misleading. It is still more honest than a bare return percentage.

In KVERX: a kverk’s Score IS a Sharpe ratio, computed from the immutable daily closes of the living record and annualised by the square root of 252 for shares or 365 for crypto. Beside it the detail page shows a Sharpe for the selected range, which — unlike the Score — changes as you move the range.

See also: Score · Drawdown · Volatility

Selling a borrowed instrument with the intention of buying it back cheaper later.

The gain is capped, because a price cannot fall below zero; the loss is not, because a price can rise indefinitely. On top of that a borrow fee is paid for every day the position stays open. A short is therefore asymmetric against the trader — and still necessary, because without it there is nothing to do in a falling market but wait.

In KVERX: only a kverk with DNA #15 set to “long and short” may go short; long only is the default. A short ties up 150 % of its value in margin and counts one-and-a-half times towards exposure, so a kverk capped at 100 % can hold shorts worth at most 66.7 % of its capital.

See also: Long · Borrow fee · Margin

The 1.15 factor by which the KVERX Market multiplies every trading cost.

A kverk trained in a harsher gym will be pleasantly surprised outside; the reverse would be a disaster. A real exchange always has extra friction a model does not capture — partial fills, brief liquidity outages, the difference between a quote and a genuinely reachable price. Fifteen percent is a cushion for what we cannot compute, and admittedly a calibration choice rather than a derivation.

In KVERX: spread, slippage, fees, margin interest, short borrow fees and funding paid are all multiplied. Income is not (funding received, dividends), nor are candle prices, nor rules such as the maintenance margin — strictness may raise costs, never lower income.

See also: Fee · Slippage · Spread

The difference between the price a strategy assumed and the price the order actually filled at.

A large order moves the price — it drains the best offers and has to take the rest at worse ones. It grows faster than one would expect: the cost follows the square root of the order’s share of normal volume, so an order a hundred times larger costs roughly a thousand times more in slippage. That is why maximum position size is not only protection against risk but also against your own costs.

In KVERX: slippage is computed from the candle’s volatility and from the order size relative to average volume over the last twenty candles, and like every cost it is multiplied by 1.15. A 5,000 USD buy in a very liquid instrument costs nine cents in slippage; a buy a hundred times larger costs 91 USD.

See also: Liquidity · Spread · Fee

Splitting a share into more units at a proportionally lower price; the value of the holding is unchanged.

A 200 USD share becomes two 100 USD shares in a 2:1 split. The owner holds the same value, only the number on the statement changes. Left untreated it looks in the data like a fifty percent crash — and every stop, every indicator and every price-derived reward would fall for it.

In KVERX: for an open position the simulator recomputes quantity, average price, stop and ATR, and does the same for pending orders. The equity curve therefore makes no artificial jump and the stop stays where the DNA put it.

See also: Dividend · Candle · Stop-loss

The difference between the price at which an instrument can be bought immediately and the price at which it can be sold immediately.

It is the first cost of every trade and it is paid before the price moves at all: buy and sell straight away and you are down exactly one spread. Spread width grows with uncertainty and shrinks with liquidity, so on a thin, choppy market it can be several times wider than usual.

In KVERX: the simulator takes the larger of two estimates — a fixed floor per universe (2–5 basis points) and an estimate from range and volume that can triple on a thin candle. Half a spread is added to each side of a trade, and the whole thing is multiplied by the 1.15 strictness factor.

See also: Bid · Ask · Liquidity

A price level at which a losing position closes automatically.

A stop is not a forecast — it is a decision made in advance, calmly, about when you will admit this trade did not work. Its one real weakness is the gap: if the market opens beyond the stop, the fill happens at the opening price and the loss is larger than planned. A stop without that admission does not work.

In KVERX: the stop is set in DNA #11 as an ATR multiple, 1× to 5×, defaulting to 2.5×. It is evaluated on every candle the position lives through, and the first touch wins. It also defines one R — the unit in which the kverk receives its marks.

See also: ATR · R-multiple · Gap

How often a kverk evaluates the situation: every five minutes, every hour, or once a day.

A faster tempo means more decisions, more trades and more costs paid — and costs are paid every time, while profit is not. A slower tempo means fewer opportunities, but also less friction and a longer wait before enough experience accumulates to learn from.

In KVERX: the tempo is chosen in DNA #27 and locks with it. The five-minute tempo is crypto-only, because share data arrives on a fifteen-minute delay and deciding every five minutes on it would make no sense. The tempo also sets the candle length the whole simulator works in.

See also: Candle · Universe · Fee

A time capsule: a real historical period a kverk races through in its own skin.

It is not a film screening. The kverk decides for itself, wins and loses for itself, receives its own marks — only day by day in a few minutes. And here is the point: two different kverks take different lessons from the same crisis, exactly like two siblings at the same school with the same teacher. The cautious one remembers “this is what the start of a fall looks like, next time I leave earlier”, the bold one “the bottom was a cheap place to buy”.

In KVERX: trading does not stop during a training — a copy of the brain trains on a temporary portfolio, and after the exam the new brain is swapped atomically between two market cycles. The swap is unconditional: an experience cannot be un-lived, and the Exam is honest feedback rather than a gate. One purchased unit = one run on one kverk, and a unit never expires.

See also: Exam · Brain maturity · Loyalty training

A market regime in which price shifts one way for a long stretch and pullbacks are shallow.

In a trend it pays to hold and it punishes selling on every wobble. The trouble is that a trend is only recognised in hindsight and never lasts — and the longer it runs, the more confidently people enter it and the more expensive its end tends to be. Momentum is a trend strategy; mean reversion is its opposite.

In KVERX: there is no trend switch. A kverk perceives one through the “momentum”, “breakout” and “relative strength” senses, and through DNA #3, which decides whether it takes profits early or lets them run.

See also: Momentum · Range · Market regime

A curated group of instruments and one accounting currency in which a kverk trades.

There are five: US shares, Europe, Asia, World (mixed) and Crypto. The universe decides which instruments the kverk can see at all, when the market is open, what a trade costs and what the maximum leverage is. It also determines which senses are available — a crypto exchange has no earnings season, and share data does not permit a five-minute tempo.

In KVERX: the universe is chosen in DNA #26 and locks with it; a different market means a new kverk, or an offspring with a low usable inheritance. Each universe accounts in a single currency (USD, EUR or USDT) and every kverk in it starts with 10,000 units — the amounts are not converted at an exchange rate, because both the Score and the return are percentages.

See also: Watch-list · Exchange · Tempo

A measure of how strongly a price fluctuates over a given period.

High volatility does not indicate a direction — it indicates uncertainty. In practice it forces wider stops, smaller positions and faster alternation of gains and losses; the same strategy therefore trades differently in a choppy market than in a calm one, without anything in it having changed. Volatility also clusters: after a wild day the next one tends to be wild too.

In KVERX: volatility is one of the fourteen senses and is measured against a calibrated normal of 0.83 % ATR per hourly candle. The DNA #14 fuse uses it as well, shrinking positions in a restless instrument — never enlarging them, not even in dead calm.

See also: ATR · Fuse · Market regime

The number of units of an instrument that actually changed hands in a given period.

Volume is the only number on a chart that measures participation rather than price. A move on heavy volume was confirmed by many people; the same move on no volume is usually noise or an empty corridor. Volume also has its own daily rhythm — there is several times more of it at the open and before the close than at midday.

In KVERX: volume feeds two senses (“volume/VWAP” and “microstructure”) and, above all, the costs — the thinner a candle is relative to normal, the wider the estimated spread, up to three times.

See also: VWAP · Liquidity · Breakout

The volume-weighted average price — what the “average” participant of a stretch paid for the instrument.

Unlike a plain average, VWAP counts each price as many times as was traded at it, so a brief spike on a handful of units does not move it. It is a reference point for large players: buying below VWAP means buying better than the average of the day.

In KVERX: the “volume/VWAP” sense measures how far above or below the twenty-candle VWAP the current price sits and multiplies that by the strength of volume. A price above VWAP on double volume therefore gives a strong signal; the same price on dried-up volume gives almost none.

See also: Volume · Sense · Liquidity

The specific list of instruments from its universe that a kverk evaluates every cycle.

A watch-list holds five to a hundred instruments; the default is the twenty most liquid. A narrower list turns a kverk into a specialist — it lets you build a “semiconductors only” kverk — and a wider one gives it more opportunities at the price of decisions being diluted through the max-positions fuse.

In KVERX: the list is part of the locked DNA (#29), so it does not change after birth; the quick path picks it automatically by liquidity, the detailed path lets you tick it yourself. A kverk applies the same network to each instrument independently, so the length of the list changes neither the shape of the brain nor the time it takes to mature.

See also: Universe · Delisting · Position

The place where you assemble your own rule from conditions over senses and indicators and hand it to a kverk as an extra input.

A rule does not return yes/no — it returns a soft number from −1 to +1 according to how many conditions hold. A condition that cannot be evaluated (missing data, too few candles) is skipped: “don’t know” must not count as “no”, because “no” would pull the result the other way.

In KVERX: the builder offers a vocabulary of every sense and indicator (RSI, MACD, ATR, VWAP…), at most twenty conditions and a safe interpreter — a rule is a data structure, not code, and is never executed. It is stored in your account library, and a kverk receives a COPY at birth, so editing or deleting it later never touches a living kverk. After thirty days an effectiveness figure appears, along with a warning for a dead rule.

See also: Sense · RSI · MACD