Model Reference

Detailed specification for every trading strategy active in the bot. Each model captures a distinct market edge.

How it all works together: the bot runs 12 models in parallel, each scanning for its specific setup. The regime allocator determines which models are active based on current market conditions (Stable / Normal / Turbulent / Crisis). The adaptive exit manager and scaled-exit risk manager handle position management differently for each model type.
๐ŸŸข Long ๐Ÿ”ด Short ๐Ÿ”„ Sweep ๐Ÿง  Dual โ†ฉ๏ธ MeanRev ใ€œ Choppy โšก Gap โš–๏ธ Pair ๐ŸŽฏ Scalper ๐ŸŒŠ Wave L ๐ŸŒŠ Wave S ๐ŸŽฏ PnD ๐Ÿšช Exit
๐ŸŸข

Long Momentum

Trend-following breakouts in bull markets
1D ยท TREND XGBoost
Captures sustained upward moves in trending instruments. Enters after a breakout above SMA50 with positive RSI, volume confirmation, and regime alignment.

How it works

  • Bar timeframe1D
  • ML classifierXGBoost
  • Entry triggerClose > SMA50, RSI 35โ€“85
  • Stop loss1.5โ€“3.0ร— ATR
  • Profit target2.5โ€“4.0ร— ATR
  • Best regimeStable / Normal
  • Max hold10 days

When it fires

Example: BTCUSDT is above its 50-day SMA, RSI is 58 (not yet overbought), and ML model predicts continuation probability of 0.62 (above 0.55 threshold).

Example trade

Entry: BTCUSDT @ $78,200 (Mar 15)
Stop: $76,300 (2.4ร— ATR)
Target: $81,500 (3.5ร— ATR)
Exit: $81,500 after 4 days โ†’ +4.2% โœ“
๐Ÿ”ด

Short Breakdown

Trend-following short entries in bear markets
1D ยท TREND XGBoost
The bearish mirror of the Long model. Enters short when price breaks below SMA50 with weakening momentum. Heavily weighted during Turbulent and Crisis regimes.

How it works

  • Bar timeframe1D
  • Entry triggerClose < SMA50, RSI 25โ€“65
  • Stop loss2.5โ€“3.5ร— ATR
  • Profit target3.0โ€“5.0ร— ATR
  • Best regimeTurbulent / Crisis
  • Allocation (Crisis)50%

When it fires

Example: AAVEUSDT closes below SMA50, RSI drops from 70 to 42, and volume surges on the down move. ML confidence 0.61 โ€” enter short.

Example trade

Entry: AAVEUSDT short @ $106
Stop: $112 (3.0ร— ATR)
Target: $91 (5.0ร— ATR)
Exit: $91 โ†’ +14.2% โœ“ (closed 4/20 @ +$556)
๐Ÿ”„

Liquidity Sweep

Reversals after wick rejections at swing highs/lows
4H ยท SMC
Smart Money Concept strategy. Detects when price briefly violates a swing high/low (stop-hunt) then reverses โ€” a sign of engineered liquidity grabs. Enters counter-trend at rejection.

How it works

  • Bar timeframe4H
  • Entry triggerWick > 2ร— body at swing level
  • Stop lossBeyond wick extreme
  • Profit target2ร— risk (R:R 1:2)
  • UniverseTop 15 pairs

When it fires

Example: SOLUSDT briefly spikes to $92 (previous swing high at $91), then closes at $88 on the 4H bar. Stop-hunt confirmed โ€” enter short at $88.
๐Ÿง 

Dual Choppy AI

Up/down prediction model for range-bound markets
4H ยท WALK-FORWARD
Two XGBoost classifiers trained separately for up-moves and down-moves within choppy markets. Only fires when both short-term direction prediction and reversion setup align.

How it works

  • Regime triggerRange-bound only
  • EntryBoth classifiers agree
  • ExitTight 1.5ร— ATR
  • Hold time4โ€“12 hours

When it fires

Example: BTC trading sideways between $77K-$79K for 3 days. Dual-up predicts 0.68, dual-down rejects. Enter long at lower range.
โ†ฉ๏ธ

Mean Reversion

Oversold bounces on strong, uptrending instruments
4H ยท RSI+BB
Buys dips on instruments in sustained uptrends. Looks for temporary pullbacks to the lower Bollinger Band or oversold RSI on otherwise strong-performing pairs.

How it works

  • Entry triggerRSI < 30, BB lower tag
  • PrerequisiteUptrend on 1D
  • Profit targetMid-BB or +2% bounce
  • Leverage3ร— (conservative)

When it fires

Example: ETHUSDT is up 20% over the month but today RSI drops to 28. Enter long at BB lower band, target middle band.
ใ€œ

Choppy Range Trader

High-frequency range fades in sideways markets
1H ยท RANGE
Identifies tight trading ranges and fades the extremes. Designed for low-volatility regimes where directional models underperform. Uses 10ร— leverage with tight stops.

How it works

  • Bar timeframe1H
  • Entry triggerPrice at range edge
  • Leverage10ร—
  • Stop lossJust beyond range

When it fires

Example: BTC trading in a 1% range ($78.5K-$79.2K) for 8 hours. Price hits $79.1K upper edge โ†’ short with tight stop at $79.3K, target $78.6K.
โšก

Gap Fade

Fades overnight/session gaps that typically fill
SESSION ยท REVERSAL
Crypto doesn't have traditional gaps, but session opens (Asia โ†’ EU โ†’ US) often create short-term imbalances. Enters contrarian when open deviates significantly from prior close.

How it works

  • TriggerSession open > 1.5% deviation
  • DirectionFade (opposite to gap)
  • TargetPrior session close
โš–๏ธ

Pair Hedge

Mean-reverting spreads between correlated instruments
4H ยท Z-SCORE
Market-neutral strategy. When two historically correlated assets diverge (e.g. ETH vs SOL spread widens 2ฯƒ from mean), long the cheap one, short the expensive one, profit when they reconverge.

How it works

  • EntrySpread z-score > 2
  • ExitZ-score back to 0
  • Market exposureNear-zero

When it fires

Example: SOL outperforms ETH by 8% in a week (normal spread is 2%). Enter long ETH, short SOL โ€” profit when spread normalizes.
๐ŸŽฏ

Microstructure Scalper

DOM-enriched 5-minute scalping on volatile instruments
5MIN ยท DOM
High-frequency scalper. Scans 100+ volatile instruments (ATR > 3%) on 5-min bars. Uses orderbook imbalance (DOM) as the primary entry signal. Targets 0.8% per trade with tight stops.

How it works

  • Bar timeframe5 minutes
  • Feature count22 (incl. DOM)
  • Target0.8% (tight)
  • Stop0.5% (tight)
  • UniverseATR > 3% (volatile)

When it fires

Example: METUSDT on 5-min chart shows strong bid support, volume spike, and ML predicts +0.8% move in next 30 minutes. Enter long.
๐ŸŒŠ

Wave Rider Long

Catches continuation of big gainers (+15% movers)
4H ยท DOM+OI+FUNDING
Rides coins that are already pumping. When a symbol has gained >15% in 24h with high volume, this model predicts whether the move continues another 5%+. Uses 28 features including exchange sentiment (funding, OI, long/short ratio).

How it works

  • Entry trigger>15% gain 24h, vol โ‰ฅ 3ร—
  • Feature count28
  • ExitTrailing 1.5ร— ATR (no fixed TP)
  • Risk per trade0.5%
  • Precision (backtest)43.8%

When it fires

Example: CHIPUSDT pumped +43% today on $620M volume, OI rising, funding still low. ML predicts 68% chance of continuation. Enter long with trailing stop.
๐ŸŒŠ

Wave Rider Short

Catches continuation of big losers (-15% dumpers)
4H ยท DOM+OI+FUNDING
The bearish twin. Shorts symbols already dumping, predicting continuation rather than catching the bottom. Best precision in Turbulent regime (49.5% backtest).

How it works

  • Entry trigger<-15% drop 24h
  • ExitTrailing 1.5ร— ATR
  • Allocation (Crisis)20%
  • Precision (backtest)49.5%

When it fires

Example: FIOUSDT drops -31% today, funding positive (longs trapped), OI declining. Enter short โ€” pattern suggests continuation to -40%+.
๐ŸŽฏ

PnD Hunter

Two-stage XGBoost that shorts pumps at exhaustion
4H ยท 2-STAGE ยท SCALED
Crypto's favorite pattern โ€” identify pump-and-dumps, wait for exhaustion signals, short the cascade. Two-stage classifier: Stage 1 identifies PnD patterns, Stage 2 times the exhaustion entry. Tracks serial offenders in a PnD history database.

How it works

  • Stage 1PnD classifier (15 features)
  • Stage 2Exhaustion timing (28 features)
  • Confirmations needed2 of 5 (red candle, DOM, etc.)
  • ExitsScaled: 30% @ 8%, 30% @ 15%, 40% trails
  • Stop2ร— ATR (tight)
  • Stage 1 precision55.0%
  • Stage 2 precision68.5%

When it fires

Example: TOKENUSDT pumped from $1 to $2.50 (+150%) on huge volume. Now showing upper wicks, first red candle, funding spiked to +0.12%, OI declining. Enter short with tight stop.

Ideal target

Serial offender tokens (pumped multiple times in 90 days)
Low-float / low-market-cap altcoins
Newer listings with wild volatility
๐Ÿšช

Exit Model (Adaptive)

Manages all open positions with regime-aware logic
60S ยท ADAPTIVE
Not a model that opens trades โ€” it manages all open positions. Runs every 60 seconds. Applies regime-specific exit logic: tighter trailing in Turbulent markets, looser in Stable. Doesn't require its own allocation.

How it works

  • Check frequencyEvery 60 seconds
  • Stable regimeLoose trailing (let winners run)
  • Normal regimeStandard 2ร— ATR trailing
  • Turbulent regimeTight 1.5ร— ATR trailing
  • Crisis regimeAggressive exit on small reversals

Partial exits

Scalper and PnD trades get scaled exits (30%/30%/40%). Other models exit all at once on stop/target hit.