- Ani, your guide
- New
Ani, the first-analysis guide. The first time you open AniQuant on trial, Ani walks you through it: you pick your working data, AQ Genesis searches for a few minutes (with a «Set 3 minutes» button) and, while it searches, she explains the wall, the discarded ones and the Log. When it finishes she takes you to the Winners’ Room, opens the first strategy in Performance and walks you through the summary, the equity and the strategy in plain words. If the search seals none, she offers to search again for longer or uses one of the Best discards. Call her back from Help → Getting started guide (Ani).
- Chance Lab
- New
Chance Lab (Capital & Risk). Can money management alone make you win? A coin, a roulette wheel or a custom game, and twelve betting systems (martingale, Paroli, D'Alembert, Fibonacci, Labouchère, Oscar's Grind, fixed fraction, Kelly…) on the same sequence of tosses, up to one million series in seconds using every core. You see who wins the race most often (and who loses it), the shape of the result, how capital evolves and one series in slow motion, with a plain-language reading. And instead of the coin you can use the trades of a strategy from your Portfolio (or a CSV): you see what each betting system does with your real edge, or with your own betting rule. The lesson: size does not create the edge, but it can multiply or destroy it.
- Shared account and margin
- New
Shared account in the Portfolio. The combined curve adds up each strategy as if it had its own money; in reality they share one account. The new view simulates that: with your capital, your limits (positions at once, per symbol, maximum open risk and margin) and each strategy's Position Size on the common capital, at each moment the trades that fit are taken, in order of confidence, and the rest are rejected with their reason and what they would have made. You see the account curve against the summed one, the real worst drawdown and who caused it, margin and risk usage over time, which strategies fall together with the others, and the list of rejected signals. With unlimited capital it gives exactly the summed curve: the difference is exactly what the sum hides.
- New
Shared account: brake, equal risk and margin calls. The drawdown brake cuts size every X% of drawdown, like the Turtles; with equal risk, every strategy risks the same % of capital to its stop; and the maintenance margin closes the lowest-priority positions at the end of the day if the account does not cover it, like a margin call. The «With and without margin» tab compares both versions with a report, and «Check margin bar by bar» finds the calls that were not visible at the daily close.
- New
Is your account robust? A «Robustness» tab in the shared account, with three tests in a few seconds. Capacity: the same account with other capitals and position limits; it says from which capital nothing is rejected any more and below which it is ruined. Sensitivity to priority: 50 random entry orders, to see how much the result depends on who enters first. Account Monte Carlo: 1,000 histories reordering stretches of its daily curve, with the typical worst drawdown and the probability of losing more than half.
- New
Shared account, out of sample: would you have chosen these strategies without knowing the future? In each period only those that were winning with what had been seen so far go into the account (all of them, or the best 3, 5 or 10), and it is compared with the same rule applied while knowing the future. With the PBO of the rule: out of 12 ways of choosing, the probability that the best one is so by luck.
- New
Overnight and intraday margin, daily loss cap and trades per day. In the Quote Manager card, the margin your broker asks to hold 1 contract overnight and the one it asks if you close before the session ends (no factory values: each broker asks its own). The shared account enters with the intraday margin, switches to the overnight one at the session close and, if it no longer fits, closes as your broker would. Plus a maximum loss per strategy and day, and two new figures: the portfolio's trades per day and its worst day. The session comes from the contract's data, in any time zone.
- Looking inside the bar
- New
Looking inside the bar with 1-minute data. When a bar hits both the take profit and the stop loss, or a Stop order fills in a bar that also hits the stop, open, high, low and close cannot tell what happened first, and until now the worst was assumed. If the market has 1-minute data, AniQuant looks at it and knows what really happened. Choose it in Settings → "Inside the bar"; it applies to the whole program: Laboratory, Genesis, Walk-Forward, Performance, Portfolio and TP/SL Map, on CPU and GPU. It corrects the back-adjustment price offset and the time offset by itself, and only uses minutes that match their bar. With normal stops it barely changes anything or slows down; with tight stops it changes a lot. Performance shows, in a new row, how many doubtful bars were resolved with the minutes and how many with the worst case, and only the truly doubtful trades are flagged as ambiguous.
- Performance
- New
Strategy tab: the strategy in plain words. The second Performance tab explains the strategy in sentences, without reading the code: each condition («An NR4 bar occurred 10 sessions ago…») with a «What is it?» for the indicator, the entry order, the exits converted to points, ticks and dollars at today’s price (with the real range of its trades), a drawing of target and stop, its last trade told and notes AniQuant writes by itself: the minimum win rate not to lose, time in the market versus buy and hold, what a stop weighs today against the capital, whether it holds up out of sample. All six exports carry those same sentences as a comment in the code header.
- New
Equity with five views: by date, detailed (bar by bar, with the open trade valued at every close: you see the dips inside each trade), by trade, each trade (one bar per trade) and the max and min of each trade. $ / % button, a hover card with the date, equity, return and drawdown at that moment, and the worst drawdown marked with its amount, duration and dates. The drawdown, in red.
- AQ Genesis
- New
Study first: new checkbox (ticked by default). Before starting, AniQuant runs its labs on the search period, without looking at the reserve, adds what it finds to the Genesis pool and rotates its exits with the Genesis one. It takes about one more minute and, in the four-market test, gave somewhat more positive strategies out of sample.
- Improved
Validation at 0 by default. In out-of-sample tests (ES and gold, 2021-2026), reserving 25% for validation took the most recent years away from the genetic engine and lowered results; without it, today’s Genesis ties the August version on the ES and wins on gold. The wall keeps its tests and the reserve remains the clean exam. It can be raised in Settings.
- Fixed
AQ Genesis validates on data it has not seen: training, validation and reserve. The wall measured each strategy on the same data it had been born from, and Walk-Forward used fixed 24 + 6 month windows: with two years of history none fitted and nothing was sealed. Also, Walk-Forward and Survival loaded the whole file and saw the reserved final stretch. Now there are three stretches, as in machine learning: the genetic engine searches in training, the wall checks in validation (Configure → «Validation»; 0 by default, see above; for example, 25% with a 20% reserve gives 60/20/20) and nobody sees the reserve until the seal. Walk-Forward windows adjust to the data and the log says where each cut falls. And Performance splits In-Sample and OOS for Genesis strategies too.
- New
AQ Genesis also searches with Stop and Limit orders. In Configure, "Entry order": the same boxes as the Lab. If you tick several, Genesis considers them all (each round tries one, in turn); if you tick only one, every strategy uses it. Each sealed strategy keeps its own all the way to the Portfolio and the exports.
- New
Smarter conditions. Calibrated thresholds: each threshold's range comes from the indicator's percentiles in your data, not from a fixed table, without looking at the reserve. Operators with sense: «for N bars» (Count(RSI(Close, 14) < 35, 3) >= 3), «changes direction» and «against its own average». And setup → trigger sequences (checkbox in Settings, ticked by default): «the setup happened in the last 3, 5 or 10 bars» and the trigger on the current bar. They are all AQL formulas: they run on the GPU and are exported to all six platforms.
- New
A more reliable wall. Minimum DSR (0.20 by default, on the same rung as PBO): a strategy whose Sharpe is explained by the luck of having tried thousands of combinations is rejected. PBO no longer comes out almost always high: it is measured with the candidates plus 60 random strategies from those tested. And the minimum Survival rises from 55 to 65. All three thresholds were calibrated on a real market against its shuffled copy.
- New
Random control in AQ Genesis. A new checkbox: when it finishes, Genesis repeats the same search (same settings and same seed) on a shuffled copy of the market for half the time and compares, per direction, how many strategies it seals on each. If the random copy seals as many as the real market, what was found is not an edge, and the Hall marks it with 🎲. It is the only test that separates a strategy that merely rides the market rise from one with a real edge.
- Improved
Genesis flags strategies that hardly ever leave the market. If a sealed strategy is in the market 75 % of the time or more, the Hall marks it in amber next to its name: in a rising market it makes money like buy-and-hold with a stop, and neither the search period nor the OOS can separate its edge from the market rise. It is not rejected; check it with AQ Market Shuffle.
- Improved
Genesis, faster and more thorough. It evaluates almost twice as many strategies per minute, uses the number of threads set in Settings and no longer proposes redundant conditions. With years of 1-minute data it no longer runs out of memory and stops within seconds. And it tests every indicator before repeating any group: in 10 minutes on 3 years of 1-minute @NQ, all 86 were tested.
- New
Bank by stretches in the Analysis viewer: a new tab tests the sealed strategies of a search again over the whole history and splits them into training, validation and reserve, with their robustness (R², days under water, positive years, Monte Carlo DD…). Searches now store where each stretch ends; older ones show the whole history and the robustness. And the viewer shows each file's date, size and whether it has sealed strategies, with the most recent first.
- New
Log to disk. A checkbox next to «Record log» (unticked by default) saves the Genesis log to a text file, with the run's settings at the top, so you can review it calmly.
- Exits, orders and session
- New
New exits. TP or SL at an indicator level (a stop at the 10-bar low, a target at the 20-bar average, a band or a pivot), target in R (with 2R, at twice the initial stop distance) and trailing with activation (it only starts once the trade has X in profit). In the Lab, in AQ Genesis and in AQL (stoploss: at Lowest(Low, 10), takeprofit: 2R, stoploss: trailing 1.5% after 2%). They run on the GPU and are exported to all six platforms.
- New
Stop and Limit at the level of an indicator: the order can be placed at the value of a moving average, a band, the N-bar high, the anchored average, a pivot or yesterday's low, with the usual distance, validity and rearm. For example, buying with a Limit at the pivot S2. In AQL: EntryOrder: limit at PivotS(2). Runs on the GPU and is exported to all six platforms. In the Lab, with the Order role in the E/S column; in AQ Genesis, each round takes a level with sense: a buy Limit at a support, a buy Stop at a resistance.
- New
TP and SL active from the next bar: new option in the Lab (Settings, «How to exit») and in Genesis. By default the target and the stop work from the fill, like an OCO order. With «the next bar» they are placed at the close of the entry bar, like TradeStation's «next bar» orders: it reproduces strategies written that way, which collect the overnight rebound instead of that afternoon's target. Genesis can try both, taking turns by round. In AQL: ExitsFrom: nextbar.
- Improved
Gaps on exits: if the market opens beyond the stop or the target, the trade exits at the open, as with a real broker: the stop worse than its level, the target better. Until now it always exited at the exact level, which was optimistic for gapped stops. Results of strategies with TP or SL change slightly, especially intraday (every session opens with a gap). Same on CPU, GPU, the TP/SL Map and Genesis.
- New
Daily limits: at most N entries a day, and stop trading for the day once closed trades lose or make an amount (in dollars per contract, with costs). In the Lab, AQ Genesis, AQL (daylimits:), the GPU and all six platforms. With daily bars they do nothing.
- New
Close at the end of the session. In the Lab and in Genesis, an option so that no trade stays overnight: it closes at the close price of the session's last bar or at the time you choose, and nothing new opens until the next session, also «Fridays only». The session comes from your data (the daily gap), with no hours to type and in any time zone. It runs on the GPU, is saved with the strategy, is written in AQL (`SessionExit: end` or `SessionExit: 15:55`) and is exported to the 6 platforms: with the platform's own session close in TradeStation, NinjaTrader and TradingView, and with the close time worked out from your data in MetaTrader 5, ProRealTime and AmiBroker. And the entry time window («only enter between such and such hours») finally reaches the exported code on all six platforms.
- Edge or market
- New
Edge or market?: new tab that compares the strategy with its «always» twin (same exits and costs, entering at any moment) per day in the market, year by year and without the best year. It tells whether its profit comes from its entry or is a gift from the market, in-sample, out-of-sample and over the full history. In the Lab, an Edge column in the ranking and a Hide market button; in AQ Genesis, rung 4 of the wall measures that same edge in training and validation.
- Indicators
- New
New indicators. Regime and momentum: KalmanTrend, KalmanSlopeT, VarianceRatio, CUSUM_Up/Dn, ConnorsRSI, TSI, KST, Coppock, CMO and the Yang-Zhang and Garman-Klass volatilities. Levels: pivots (PivotP, PivotS/R 1-3), the previous period's high, low, open and close, the opening range, overnight and intraday return and the anchored average (AnchoredLow/High). Structure: SwingHigh/Low, MarketStructure and the regression channel (LinRegLine/Upper/Lower). And ValueChart in four variants. In the Lab, Genesis, the GPU and all six platforms.
- New
Trend lines the way a chartist draws them (TrendLineUp, TrendLineDn and their channels TrendChannelUp and TrendChannelDn): one watched line, anchored at the major low (or high) up to 500 bars back, with no bar piercing it. 3 touches confirm it and it dies when a close crosses it: that is the breakout. On the Chart it is drawn from its anchor, with the touches and the breakout. Buying the breakout on the close on 4-hour charts beat «always in» out of sample on @NQ, @ES and @GC. In the Lab, Genesis, the GPU and the six platforms. Plus RSI and MACD divergences, horizontal support and resistance and the breakout of eight chart patterns (double bottom and top, head and shoulders and its inverse, triangles).
- Fixed
Parabolic SAR as Wilder’s: it reverses when price touches the SAR (before, only when crossing it) and, when reversing, the new SAR does not sit inside the last two bars. The same in the engine, the GPU and all six platforms, and checked against TA-Lib.
- GPU
- Improved
More searches on the GPU. Stop and Limit orders, the trailing stop, break-even, exit after N bars and the entry time window now run on the graphics card, as fast as at market. Until now they forced a CPU search, several times slower. Same trades as the CPU, to the cent.
- Fixed
GPU and CPU, identical trade by trade. In forex and stocks, with prices like 1.10235, the card could not tell apart an average and a close that were almost equal, nor the exact TP/SL level (in EURUSD, more than half the strategies came out with a different number of trades), and with a TP or SL like 0.2% some level landed on the next tick. Now they match.
- New
You can now see whether the search runs on the GPU or the CPU, and why. The Lab, Genesis and Walk-Forward show the real engine, with the card name ("GPU (NVIDIA GeForce RTX 3080)") or the reason for using the CPU ("the trailing stop only runs on the CPU").
- Export
- Fixed
TP and SL on export: TradeStation, MetaTrader 5, ProRealTime and TradingView always wrote the target and the stop in %, even when they were in $ or ATR in AniQuant (a $715 TP came out as 715 %). Now they use their own unit. Also, in TradeStation and TradingView the TP and SL did not work on the entry bar (they were placed at that bar's close), while AniQuant activates them from the fill: now they work from the fill, as in AniQuant, and with the new option, from the next bar on all six platforms. In TradingView the trailing now starts from the entry, as in AniQuant, and in MetaTrader 5 a strategy whose only self-computed indicator is the anchored average now compiles.
- Position Size
- New
More realistic Position Size. «With these numbers»: a live calculation says how many contracts the method asks for and how many remain after the minimum and whole units, and warns if with your capital the result would match Normalized (with 1,000 $ per position, the ES asks for 0.003 contracts). And the broker margin, separate from exposure: it decides how many contracts you can open, with Maximum margin to cap the % of the account tied up and a warning if any trade needed more margin than the balance.
- Other improvements and fixes
- Improved
Every notice and question, in AniQuant style. No more small grey Windows box: the program’s 180-plus notices, confirmations and name prompts now appear in AniQuant’s own dialog, in your language, with an icon for what it is (information, question, warning or error), Enter for the blue button and Esc to cancel. When naming a Walk-Forward study, a portfolio, a Survival collection or a Laboratory template, it warns if the name is empty (or, in the Portfolio, already exists).
- Improved
Walk-Forward says why it can’t run. It used to show a generic message. Now a card explains the reason: not a single cycle fits («your data covers 22 months and each cycle needs IS 36 + OOS 6»), it can’t find the strategy’s data, or the template is unusable. If nothing fits, a button sets IS and OOS months that do fit (about five cycles) and redraws the diagram. And for a strategy without an E-… code, the suggested study name is «WF» with the date.
- Improved
«No stop» warning next to the drawdown. A strategy without a stop loss holds open losses until the target and shows «100 % winners, drawdown 0»: the drawdown only counts closed trades. The Lab, Performance, Portfolio and Walk-Forward flag it wherever the drawdown is shown.
- Fixed
ATR-based TP, SL and break-even no longer look into the future. They were computed with the ATR of the same bar being checked, which includes that bar's own range: a large bar pushed the stop away just when it was threatening it. They now use the previous bar's ATR, as trading platforms do. Strategies with ATR-based TP/SL or break-even may give slightly different results.
- Fixed
Exact tick levels in forex, stocks and gold. With ticks such as 0.0001, 0.01 or 0.1, the computed level could end up a millionth above the real price, and a stop touched exactly at its level sometimes did not trigger. And the strategy name carries every decimal of the threshold (0.0012 used to show as 0).
- Fixed
The Lab grid and Performance match when part of the data is held out. A trade opened at the end of the search segment and closed in the held-out one counted in Performance's In-Sample column but not in the grid. Each trade now counts in the segment where it closes.
- Fixed
Daily reference series: same-day data. A daily series imported without a time (for example the Fear & Greed index from a CSV) was seen one session late against a daily market that closes before midnight, such as the SPX: a signal on the 19th bought on the 23rd instead of the 20th. Now, between two daily series, each day uses that day's value, as other platforms do. Reported by Oscar Cagigas (www.onda4.com).
- New
TP/SL Map: amounts in dollars using the point value (they were points without saying so), the Heatmap marks the best and worst combination and Robustness measures on net profit. And the new By period view: the best and worst combination of each year and each month, and the % years + metric, which highlights the areas that win almost every year.
- New
"Sweep settings" window. In the Lab, what you set once (the random and genetic numbers, how each order is placed, the session, maximum results) moves to a dialog with an explanation next to each setting. The panel keeps a line that sums it up.
- Improved
More readable tables in every window. Brighter, bold column headers.