AlgoGMI/ Docs
Dashboard

How to use AlgoGMI

Everything you need to create, test, and deploy trading strategies using AI.

Overview

AlgoGMI is an AI-powered strategy builder for crypto trading. Describe your trading idea in plain English and AI converts it into a backtestable strategy with real market data, fees, and slippage.

The workflow

1. Describe — type your strategy idea or pick from Discover

2. Backtest — AI generates the config, runs it against 12 months of real data

3. Analyze — see monthly breakdown, AI diagnoses strengths and weaknesses

4. Improve — refine your prompt or use AI auto-improve

5. Deploy — once Grade A or B, deploy to paper or live on Mudrex

Creating strategies

1. Natural language (AI Builder)

Describe what you want. AI understands trading concepts and converts them into precise indicator configs, entry/exit rules, and risk parameters.

"Buy BTC when momentum is strong and volume is high, 
risk 2% per trade, stop at 3%"

"Mean reversion on ETH: buy when price is 2 standard 
deviations below the mean and RSI shows oversold"

"Ichimoku cloud breakout on SOL with volume confirmation"

"SOL momentum: EMA 21 above EMA 55, RSI 50-70, 
2% risk, 3% SL, 6% TP, 1h timeframe"

2. Visual Builder (Free)

Step-by-step configuration of indicators, entry conditions, exit rules, and risk management. No AI credits used.

3. Discover tab

8 pre-built strategies with proven backtest results. Click "Build This with AI" to load any of them and customize.

Refining strategies

After creating a strategy, use the Refine tab to modify without starting over:

"Make it long-only"
"Switch to 4h timeframe"
"Add a volume filter"
"Tighten the stop loss to 2%"
"Use trailing stop instead of fixed TP"
"Exit at BB midline instead of fixed TP"

Indicator reference

AlgoGMI supports 40+ indicators. Each creates columns you can use in entry/exit conditions.

Trend indicators

IndicatorParametersColumn namesDescription
emaperiod (default: 21)ema_21Exponential Moving Average
smaperiod (default: 50)sma_50Simple Moving Average
demaperiod (default: 21)dema_21Double EMA — reduced lag
temaperiod (default: 21)tema_21Triple EMA — minimal lag
hull_maperiod (default: 20)hull_20Hull Moving Average — fastest, least lag
vwapnonevwapVolume Weighted Average Price
supertrendperiod, multipliersupertrend_10, supertrend_dir_10Trend direction. dir=1 bullish, dir=-1 bearish
ichimokutenkan, kijun, senkou_bichimoku_tenkan, ichimoku_kijun, ichimoku_senkou_a, ichimoku_senkou_bIchimoku Cloud — full system

Momentum indicators

IndicatorParametersColumn namesDescription
rsiperiod (default: 14)rsi or rsi_14Relative Strength Index — 0 to 100
macdfast, slow, signalmacd, macd_signal, macd_histMACD line, signal, and histogram
stoch_rsirsi_period, stoch_period, k, dstoch_rsi_k, stoch_rsi_dStochastic RSI
stochastick_period, d_periodstoch_k, stoch_dStochastic oscillator
cciperiod (default: 20)cci_20Commodity Channel Index
mfiperiod (default: 14)mfi_14Money Flow Index — volume-weighted RSI
rocperiodroc_14Rate of Change — % price change over N bars
williams_rperiodwilliams_r_14Williams %R — -100 to 0

Volatility indicators

IndicatorParametersColumn namesDescription
bbandsperiod, std_devbb_upper, bb_lower, bb_middle, bb_widthBollinger Bands
atrperiod (default: 14)atr_14 or atrAverage True Range — volatility in price units
keltnerema_period, atr_period, multiplierkeltner_upper, keltner_lower, keltner_middleKeltner Channel
donchianperiod (default: 20)donchian_upper_20, donchian_lower_20, donchian_middle_20Donchian Channel — highest high / lowest low
historical_volperiodhvol_20Annualized historical volatility %

Volume indicators

IndicatorParametersColumn namesDescription
volume_smaperiod (default: 20)volume_sma_20Volume moving average for spike detection
obvperiod for SMAobv, obv_smaOn-Balance Volume
cmfperiod (default: 20)cmf_20Chaikin Money Flow

Other indicators

IndicatorParametersColumn namesDescription
adxperiod (default: 14)adx_14 or adxTrend strength — 0 to 100
pivot_pointsnonepivot, pivot_r1, pivot_s1, pivot_r2, pivot_s2Classic pivot points from previous bar
prev_daynoneprev_day_high, prev_day_low, prev_day_closePrevious day OHLC — works on any timeframe
candle_patternsnoneis_doji, is_hammer, is_shooting_star, is_bull_engulf, is_bear_engulf, is_big_candle, is_inside_barCandle pattern detection (1 or 0)

Statistical / Quant

IndicatorParametersColumn namesDescription
zscoreperiod (default: 20)zscore_20Standard deviations from rolling mean
percentileperiod (default: 100)pctrank_100Price percentile rank — 0 to 100
linreg_slopeperiod (default: 20)slope_20Linear regression slope — normalized % per bar
correlationperiodcorr_price_vol_20Price-volume correlation
volatility_ratioshort, longvol_ratioShort vol / long vol. >1 = expanding
returnsperiodreturn_14N-bar return as percentage
range_pctperiodrange_pct, avg_range_20Bar range as % of close
distance_from_highperioddist_high_20Distance from N-period high as %
distance_from_lowperioddist_low_20Distance from N-period low as %

Lookback / Pattern

IndicatorParametersColumn namesDescription
new_highperiodnew_high_201.0 if close is at N-period high
new_lowperiodnew_low_201.0 if close is at N-period low
bars_since_highperiodbars_since_high_20Bars since N-period high
consecutive_upnoneconsec_upCount of consecutive up closes
consecutive_downnoneconsec_downCount of consecutive down closes
rsi_divergencersi_period, lookbackrsi_bull_div, rsi_bear_divRSI divergence detection (1 or 0)

Time-based columns (always available)

IndicatorParametersColumn namesDescription
(auto)nonehour, minute, day_of_week, monthTime components from candle timestamp
(auto)noneis_asia, is_london, is_newyork, is_overlapTrading session flags (UTC)
(auto)noneis_london_killzone, is_ny_killzone, is_asian_killzoneICT-style kill zone flags
(auto)noneis_weekend, is_weekday, is_monday, is_fridayDay flags

Writing conditions

Conditions use column names, operators, and values. Multiple conditions are joined with AND or OR.

Operators

>    greater than         ema_9 > ema_21
<    less than            rsi < 30
>=   greater or equal     adx_14 >= 25
<=   less or equal        close <= bb_lower
==   equals               supertrend_dir_10 == 1

Combining conditions

AND — all must be true:
  ema_9 > ema_21 AND rsi > 50 AND adx_14 > 25

OR — any group can trigger:
  (close < bb_lower AND rsi < 30) OR (is_hammer > 0 AND rsi < 40)

Math expressions:
  volume > volume_sma_20 * 1.5
  close > ema_21 + atr_14 * 0.5

Time-based filters:
  is_london == 1 AND rsi < 30
  is_ny_killzone == 1 AND macd_hist > 0

Exit strategies

AlgoGMI supports multiple exit types. The AI automatically picks the best approach for your strategy, or you can specify.

Fixed SL/TP

Set a fixed stop loss and take profit percentage. The most common approach.

"exit_rules": {
  "stop_loss_pct": 3.0,
  "take_profit_pct": 6.0
}

Dynamic TP (BB midline)

For mean reversion strategies, use Bollinger Band midline as a dynamic take profit target. The TP level adjusts with the market.

"exit_rules": {
  "stop_loss_pct": 2.0,
  "take_profit_target": "bb_middle"
}

Trailing stop

Locks in profits as the trade moves in your favor. Best for trend-following strategies.

"exit_rules": {
  "stop_loss_pct": 3.0,
  "trailing_stop_pct": 2.5
}

Trailing + cap

Trailing stop with a maximum take profit. Captures trends while capping upside.

"exit_rules": {
  "stop_loss_pct": 3.0,
  "trailing_stop_pct": 2.5,
  "take_profit_pct": 10.0
}

Cooldown

Wait N bars after closing a trade before entering a new one. Prevents revenge trading.

"exit_rules": {
  "cooldown_bars": 3
}
The AI automatically tests multiple exit modes during backtest optimization and picks the best one.

Backtesting

Backtests run against real Bybit historical data with realistic cost assumptions.

What's included

Real data — actual OHLCV candles from Bybit

1-minute resolution exits — SL/TP checked on 1m candles for precision

Taker fees — 0.04% per side

Slippage — 0.03% per side

Funding rates — 0.01% per 8h for open positions

Position sizing — risk-based (% of capital per trade)

Grading system

Grade A — PF >= 1.8, Sharpe >= 1.0, WR >= 45%, DD < 15%  -> Deploy ready
Grade B — PF >= 1.3, Sharpe >= 0.5, WR >= 40%, Return > 1%  -> Deploy or improve
Grade C — PF >= 1.0, Return >= 0%  -> Marginal, improve first
Grade D — PF >= 0.7  -> Unprofitable
Grade F — PF < 0.7  -> Broken, try different approach

AI improvement

AI analyzes your backtest results and suggests targeted improvements. You get 3 AI improvement attempts per strategy.

Weakest month analysis — AI identifies which months underperform and why

Specific recommendations — not generic tips, but exact parameter changes

Auto-improve — AI generates a surgically improved version

Deployment

Deploy strategies to paper or live trading on Mudrex.

Paper trading

Simulated trading with virtual capital. No API keys needed. Free for all users. Runs 24/7 with real market data from Bybit.

Live trading

Real orders placed on Mudrex Futures. Requires Mudrex API keys (trade-only permissions, no withdrawal). Uses the same indicator engine and signal evaluation as backtesting — what you see in backtest is what runs live.

How live execution works

• Every candle close: fetch data → compute indicators → detect regime → evaluate conditions → place order

• Kill switch: auto-stops if drawdown exceeds your limit

• Cooldown: configurable wait period between trades

• All 40+ indicators work identically in live and backtest

Risks & warnings

Trading cryptocurrencies carries significant risk of loss. Past backtest performance does not guarantee future results. AlgoGMI is a tool — it does not provide financial advice.

Backtesting limitations

Overfitting — a strategy that works perfectly on history may fail live. Watch for: Sharpe > 3, win rate > 80%, fewer than 30 trades.

Survivorship bias — only currently listed pairs are in the data.

Slippage underestimate — real slippage during extreme volatility can be worse.

Risk management rules

• Never risk more than 2% of capital per trade

• Always use a stop loss

• Paper trade for at least 2 weeks before going live

• Start with small capital on live

• Set a max drawdown kill switch of 15-20%

Strategy examples

Proven patterns you can use as starting points or type directly into the AI builder.

1. Trend following

"BTC trend following: buy when Supertrend is bullish 
and MACD histogram is positive and ADX > 25. 
2% risk, 3% stop, 6% TP. 4h timeframe."

2. Mean reversion

"ETH mean reversion: buy when close < BB lower 
and RSI < 30. Exit at BB midline. 2% risk, 2.5% SL. 1h."

3. Volume breakout

"SOL breakout: buy when close > donchian upper 20 
and volume > volume SMA 20 * 1.5 and ADX > 25. 
2% risk, 3% SL, trailing stop 3%. 1h."

4. Momentum + session filter

"BTC momentum during London session: buy when 
EMA 9 > EMA 21 and RSI > 55 and is_london == 1. 
2% risk, 2% SL, 4% TP. 15m."

5. Multi-indicator confluence

"AVAX: buy when Ichimoku cloud breakout 
and MACD crossover and volume surge. 
2% risk, trailing stop 3%. 1h."

Coming soon

These features are planned but not yet available.

Telegram trade alerts — get notified when your bot opens/closes trades

Strategy sharing — public URLs to share your strategies with others

Equity curve visualization — interactive equity chart in agent detail panel

Multi-coin deployment — run one strategy across multiple pairs simultaneously

CoinDCX support — additional exchange support beyond Mudrex

Custom indicator formulas — define your own calculated columns

API access — programmatic access to create, backtest, and deploy strategies