Which exit settings work best?
Backtests the strategy of the left panel (its symbol, period, costs and fills) once for every combination of the values you give below, ranks the combinations on the first part of the period and checks them on the last part, which was not used to choose.
Settings to vary
Tick a setting and give the values to try, separated by commas, or as from:to:step (0.5:3:0.5 = 0.5, 1 … 3). “and off” also tries it switched off. Every combination is one full backtest.
Previous sweeps
Map of the combinations
Combination
All combinations
Click a column to sort, a row to see itRun a strategy live
Set up a strategy in the Backtest tab, then start it here. It trades a virtual portfolio on the real-time market. Each verification reruns it on the data it recorded and shows the first step where the live run and the backtest differ, with the events around it.
Live vs backtest
Market data feeds
Binance public websockets, recorded into the storePrice, indicators & trades
Refreshed every 15 s · since trading startedThe chart starts with the first trading candle.
Virtual portfolio
Events
Feed connections, missed and late trades, restarts, timing — the same lines as in the journalClosed trades
Test technical analyses over a period
Add one or more technical analyses on the left and set them as you like. Every signal is drawn on the price: a gray zone from the signal until the target was reached, a red zone until the loss limit was hit. Change a setting and the chart follows.
Signals on the price
Month by month
Win rate of the signals decided (timeouts left out); the line is random entriesSignals
Click a signal to see it on the chart with its target and loss limit.
See what market filters exclude
Add filters on the left (volume, activity, volatility, hours, any indicator…) and set them: the chart shows in red the moments they exclude, and how flat the market was there.
Excluded moments on the price
Below the price: the volume, then each filter's measure with its limits (dashed).
Each filter
Separation: share of the flat moments excluded minus share of the moving moments excluded (0 = like excluding at random, 1 = every flat moment and no other). The time to move is the median.
When is the market flat?
flat momentsexcludedBy hour of the day (UTC) and by weekday, over the period: the share of flat moments, and the share the filters exclude.
Excluded periods
Click a period to see it on the chart.
Nothing is excluded.
Market data
Exports for analysis
kept 7 days on their serverLive recordings
From every technical analysis to a tested strategy
Pick a symbol, a long period and a threshold, then run. The period is cut in three: the first part chooses everything (readings, filters, exits, model), the second checks each choice, and the last is only backtested at the end. A strategy that still makes money there has passed a fair test; one that does not was fitted to the past.
Steps
Backtests
The test period was used for nothing else1. Screening
Regimes
Where the best readings change2. Filters
3. Screening with the filters
4. Strategy
5. Exits
6. Model
A model to filter a strategy's entries
Pick a rule-based strategy and a long period, then train. The model learns which of the strategy's signals tend to win from what the market looked like when they fired. You then see whether it really helps on a period it never saw, with real backtests of the strategy with and without it. Most of the time, on real markets, the honest answer is a small edge or none: this tab tells you which.
Holdout: the strategy with and without the model
Threshold: which signals to keep
Walk-forward
Each part predicted by a model trained on the parts before itWhat the model looks at
Jobs
Find the signals that work
Pick a period and timeframes, then run. Every technical analysis is tested with its default settings. So is the order-flow detector. Each signal is a success if the price reaches the target before the loss limit. The best signals are shown on the price chart, with green where they worked and red where they failed, and a breakdown by market condition.
Final strategy
The most reliable signals
Ranked on the discovery period by win rate, penalised for small samples (95% lower bound)Signal groups: analyses that confirm each other
Settings: which values work better?
Flat markets: which filter keeps the entries out?
On the market (random entries at every minute)
On the best readings
When does it work?
Win rate in each market condition, next to random entries in the same condition and on the same sides. The whole period is used.
All candidates
Backtest a strategy
Pick a strategy and its settings, choose the market data and fill model, then run. Results show the price chart with indicators and trades, the simulated portfolio against buy & hold, performance metrics and every trade.
Price, indicators & trades
Portfolio (simulated)
Marked to market at every candle close · fees includedDetections
Price move after each detection, in the detected direction (positive = the price went the way the detection said). Compare with the typical move to judge whether the signal adds anything.
Compare variants
Backtests of the same period with one setting changed each, compared side by sideMeasures whether following a detection pays after costs (from the price you would have got, not the detection price), and what each setting from the review changes: the order book (live sessions never have it), the fast normal, where the trend is judged, and a pause after a loss.