Algorithmic Trading
for LMEX Exchange

Trend-following strategies exploit the statistical tendency for assets to continue moving in one direction. EMA crossovers detect momentum shifts — when the fast line crosses the slow, market consensus has shifted. Supertrend uses ATR to set adaptive trailing stops that widen during high volatility and tighten during calm periods. Donchian breakouts capture the "new high = continuation" pattern popularised by Turtle Traders. MACD histogram measures acceleration of a trend, not just direction — ideal for timing entries within an established move.

Risk 1–2% per trade. Formula: Qty = (Account × 0.01) ÷ (Entry − Stop). For EMA cross, stop below the most recent swing low (longs). For Supertrend, the band IS the stop — size accordingly. Scale into winners: add a second unit when the trade moves 1R in your favour. Max 3 units. Reduce size 50% if the last 5 trades had fewer than 40% winners.

Whipsaw is the primary risk in flat choppy markets. Apply ADX > 25 filter before taking signals. Trend strategies inherently have 40–55% win rates compensated by large R:R (1:2+) — never risk more than 2%. Gap risk on daily bars can blow through stops. Three consecutive losses may signal a regime change — pause and reassess.

Mean reversion exploits the statistical tendency for price to return to a historical average after extreme deviations. Bollinger Bands define the normal range as ±2 standard deviations — roughly 95% of price action stays within. RSI divergence catches hidden momentum shifts: price makes a new low but RSI makes a higher low, revealing exhausted selling pressure. Z-score analysis quantifies exactly how statistically extreme a price level is, allowing consistent entry thresholds across any asset.

Risk 0.5–1% per trade — smaller than trend strategies because reversions can fail and become trends. Size = (Account × 0.005) ÷ (Entry − Stop). BB stop: above upper band on short, below lower band on long. Consider scaling in: take 50% at first signal, add 50% if price extends further (improving average). Hard stop at Z = ±3.5.

In trending markets price walks the bands — touching the lower BB repeatedly without reversing. Only take reversion trades when ADX < 20 and price is range-bound. RSI can stay oversold for extended periods during strong trends. Never average down beyond 2 adds. Three consecutive losers beyond plan — assume trend regime and pause.

Momentum strategies are built on the empirical observation that assets moving strongly in one direction tend to continue. Rate of Change (ROC) quantifies this as a pure percentage — no smoothing, no lag — making it honest about how much ground an asset has covered. Cross-asset momentum ranks all pairs by recent return and longs the top performers. StochRSI applies the stochastic oscillator to RSI itself, making it more sensitive — ideal as a precise entry trigger within an established trend rather than as a standalone system.

Single-asset momentum: risk 1–1.5% per trade, stop at the swing low predating the momentum move. Cross-asset portfolio: equal-weight allocations, never exceed 5% per position. Rebalance weekly. If ROC > 2× the asset's average ROC, reduce size by 25% — momentum is already extended. StochRSI stops: just below the crossover candle's low for longs.

Momentum crashes are violent and sudden — when crowded momentum trades reverse, they reverse hard. Always maintain a hard stop. Momentum strategies underperform dramatically in ranging markets — use ADX or trend filter to disable signals when ADX < 20. Factor crowding: when many algos buy the same top-ranked assets, correlations spike and diversification disappears.

Grid trading profits purely from price oscillation — no directional prediction required. The bot places buys below current price and sells above at fixed intervals. When a buy fills, it places a sell one level higher; when that fills, another buy one level lower. Each round trip earns one grid interval as profit. The strategy profits from any movement — up or down — as long as price stays within the defined range. Geometric grids space levels equally in percentage terms (better for volatile assets); arithmetic grids in dollar terms.

Total USDT ÷ Number of grids = USDT per level. Minimum per level must exceed $5 (LMEX minimum). Recommended: at least 20 grid levels, budget ≥ $200. Capital reserve rule: hold enough to fill ALL buy orders simultaneously if price drops to range bottom. Formula: Reserve = num_grids × qty_per_level × bottom_price. Start with ≤ 5% of total capital.

Catastrophic risk: price breaks out of range — you end up holding a full directional position at the worst level. Always set a hard stop at the bottom of the grid range. Define range using 3–6 months of historical price data. Avoid deploying grids when BB squeeze signals a breakout is imminent. Review and reset the range at least monthly.

Market making profits from the bid-ask spread on every completed round trip. The bot quotes a bid below mid-price and an ask above — when both fill, the difference minus fees is profit. Inventory skew is critical: as the bot accumulates a directional position it adjusts quotes, widening the spread on the heavy side to discourage further accumulation. On LMEX, the maker fee rebate means every filled limit order generates positive economics — earning from the exchange itself while capturing the spread.

Quote size = 0.1–0.5% of capital per side. Start extremely small — infrastructure issues cause double-fills or missed cancellations. Max inventory: 1% of total capital. Hard inventory limit: if position reaches max, cancel all quotes on the heavy side immediately. Spread minimum must exceed 2 × taker fee (0.2% on LMEX). Recommended spread: 10–20 bps.

Adverse selection — informed traders consistently trade against quotes, leaving losing inventory. If fills are consistently profitable for the counterparty, you're being adversely selected. Inventory risk: directional moves generate mark-to-market losses. Latency risk: slow cancel-replace means stale quotes fill at unfavourable prices after news. Quote refresh rate < 500ms recommended. Emergency shutdown: cancel all orders first, inventory second.

Dollar-cost averaging removes the hardest problem in trading: timing. Purchasing a fixed dollar amount at regular intervals automatically buys more units when price is low and fewer when high — lowering average cost over time without prediction. The intelligent DCA layer adds RSI conditioning: when an asset is statistically oversold (RSI < 30), the bot increases purchase size by 2–3×, concentrating capital at the best entry points. Particularly powerful during bear markets when emotional traders are selling.

Base purchase = fixed % of monthly discretionary budget. Never exceed 10% of total portfolio in a single DCA campaign. RSI multipliers: 3× at RSI < 25, 2× at RSI < 35, 1× at RSI 35–65, 0.5× at RSI > 65, 0× at RSI > 75. Maximum total exposure cap: never hold more than 20% of portfolio in any single asset regardless of RSI. Track cost basis carefully.

DCA does not protect against a fundamentally broken asset. Apply only to high-conviction, liquid assets (BTC, ETH on LMEX). The strategy can create significant unrealised losses before recovering — ensure months-to-years runway to withstand drawdowns. Never use leverage in a DCA strategy. Maximum total loss threshold: if the position is down 60% from inception, pause and reassess the thesis.

Statistical arbitrage exploits the mean-reverting relationship between two historically correlated assets. When their spread deviates beyond a statistically significant threshold (Z-score), the trader goes long the underperformer and short the outperformer — betting the spread compresses back to its historical mean. The foundation is cointegration, stronger than correlation: two cointegrated series may diverge temporarily but are mathematically bound to reconverge. ADF test confirms this. Natural LMEX pairs: BTC-PERP/ETH-PERP, ETH-PERP/SOL-PERP, OIL-PERP/BRENT-PERP.

Position legs in equal dollar value, not equal quantity. Use the hedge ratio (β from OLS regression) to adjust relative sizes. Never risk more than 2% per spread trade. Entry at Z = ±2.0, exit target at Z = 0, hard stop at Z = ±3.5. Keep both legs on ISOLATED margin on LMEX perps. Recalibrate the hedge ratio monthly.

Correlation breakdown is the catastrophic failure mode — cointegrated assets can decouple permanently due to protocol changes or regulatory events. Re-run ADF test weekly; if p-value > 0.10 close immediately. Both legs moving against simultaneously is possible. Hard time stop: if the trade hasn't converged within 14 days, close regardless of Z-score.

• Risk ≤2% of account per trade
• Set stop loss BEFORE entering
• Use ISOLATED margin on futures
• Halt bot at 10% drawdown
• Log every trade with full context
• Test on paper first — always
• Keep API keys IP-whitelisted

• No stop loss on leveraged positions
• Add to losing positions (averaging down)
• Use CROSSED margin with multiple bots
• Deploy untested code with real funds
• Give bots withdraw permissions
• Trade max leverage (>10x for bots)
• Ignore drawdown — cut the bot

Scalping extracts tiny profits from microstructure inefficiencies at high frequency. Order flow scalping reads the live tape to detect aggressive buyers or sellers — when market buy orders dominate for a sustained window, price is about to be pushed up. This predicts short-term direction before candlestick patterns confirm it. The EMA ribbon (8/13/21 on 5m) filters for alignment: when all three EMAs are stacked bullish AND flow is bullish, the probability of a 3–5 tick gain is significantly elevated. Tape reading patterns (icebergs, absorption, thin walls) add conviction.

Extremely small per trade — 0.25% risk maximum. At 0.1% taker fee per side, a 0.5% target means 40% of gross profit goes to fees. Use post-only limit orders wherever possible to earn maker rebate. Size = (Account × 0.0025) ÷ (Entry − Stop). Stop: 1–1.5 × ATR(14) on 1m chart. Target: 1.5–2 × stop. If 5 consecutive stops hit, halt for 30 minutes.

Fee erosion is slow death for scalpers — calculate fee-adjusted expectancy before scaling up. Emotional degradation after losses leads to over-trading and revenge trading. Use a strict daily loss limit (2% of account). Slippage on fast markets can turn a 3-tick target into a scratch trade — account for 0.5–1 tick slippage in backtests. Sub-100ms execution and stable internet required.

Funding rate arbitrage captures the periodic payment that perpetual futures make to balance the market. When funding is positive (longs pay shorts every 8h), holding short futures + long spot in equal dollar value creates a delta-neutral position — price moves cancel across legs — while funding payments arrive as pure income. On LMEX, OIL-PERP and BRENT-PERP often carry elevated funding during commodity moves. The yield is calculable in advance (8h rate × 3 × 365 = APR), making it one of the most transparent strategies available.

Equal dollar value on both legs — non-negotiable for delta neutrality. Size based on targeted yield: calculate the funding APR at entry, compare to capital deployed. Scale up only when basis (futures − spot price) is confirmed stable. Max exposure: 20% of total capital per funding arb position. Use ISOLATED margin on the futures leg — never cross margin.

Not truly risk-free. Key risks: (1) Leg slippage on open/close — enter with limit orders, accept slightly worse entry rather than chasing. (2) Funding flip — if sentiment shifts, funding goes negative and you're paying. Monitor every 8h. (3) Basis divergence during extreme volatility can trigger margin calls on the futures leg even though overall position is neutral. Keep 30% of futures leg as maintenance buffer.

VWAP represents the true intraday fair value of an asset — where the most volume has traded. Institutional algorithms anchor to VWAP for performance benchmarking, meaning large players consistently buy pullbacks to VWAP in uptrends and sell rallies in downtrends. This creates self-fulfilling support and resistance. TWAP is an execution algorithm, not a signal — it solves the large order problem by slicing into equal-sized chunks across a defined time window, minimising market impact and achieving a price close to the arithmetic mean of the period.

VWAP signal: risk 1–1.5% per trade. Stop below VWAP for longs — price breaking VWAP invalidates the thesis. Size = (Account × 0.01) ÷ (Entry − VWAP). TWAP execution: divide total intended position by number of slices. Rule of thumb: 1 slice per 10% of target time window. Monitor fill quality — if market is moving strongly, adjust slice frequency.

VWAP resets at midnight — loses relevance for multi-day positions. Institutional players know where VWAP is and can manipulate price around it to flush stops. Use a 0.2–0.3% buffer below VWAP for stops rather than exactly at VWAP. TWAP risk: if market moves strongly during execution, average fill price degrades. Set a price limit on TWAP slices.

Order flow analysis reads the live microstructure of the market — actual buy/sell transactions as they happen — rather than waiting for candlestick patterns to form. Order Book Imbalance (OBI) measures whether resting bids or asks dominate the visible book. Cumulative Delta tracks the running total of aggressor volume (market buys minus market sells) — persistent positive delta in an uptrend confirms institutional buying; divergence warns of hidden selling. Large wall detection identifies unusual concentrations of resting limit orders signalling institutional interest levels.

Very tight stops and small sizes — OBI signals are short-lived (seconds to minutes). Risk 0.5% per trade maximum. Stop placed just beyond the imbalance level. Target = 1.5–2× stop distance. Signal confidence scales size: OBI > 0.5 = full 0.5% risk; OBI 0.3–0.5 = half size. Never hold through a funding reset or major scheduled event. Use limit orders to enter.

Spoofing is the primary risk — large resting orders placed with intent to cancel as price approaches. Watch whether walls hold as price approaches or disappear. HFT noise on sub-1m timeframes makes OBI signals extremely noisy — stick to 1m minimum. Cumulative delta divergences can persist for extended periods in trending markets. Always require price action confirmation before acting on OBI signals alone.

The Wyckoff Method describes how institutional smart money systematically accumulates or distributes large positions without moving the market against themselves. During accumulation, price is pushed into a range, retail traders are shaken out at the Spring (a brief dip below support), then markup begins. The Spring is the highest-conviction entry — a false breakdown that immediately recovers, trapping short sellers. Distribution mirrors this: an Upthrust After Distribution (UTAD) breaks above resistance briefly, trapping breakout buyers, then reverses. Volume confirms — accumulation Springs are accompanied by high volume (institutional buying) and immediate recovery.

Spring entries: full risk 2–3%. Other Wyckoff events (AR, ST, SOS): 1–1.5% risk. Stop for Spring longs: below the Spring low by 0.5 × ATR. Target for markup: measure the depth of the accumulation range and project that distance above range resistance. Scale out 50% at 1R, move stop to breakeven, let remainder run.

Wyckoff counting is subjective — practitioners often disagree in real-time. Hard rules (Wave 2 cannot exceed Wave 1 start) help but alternation guideline creates ambiguity. A Spring that fails to recover means the accumulation was actually continuation — exit immediately at stop, no averaging down. Crypto markets can compress or skip phases due to 24/7 trading. Phase identification improves significantly with practice.

Elliott Wave Theory describes market movement as alternating 5-wave impulse + 3-wave correction cycles. Rules are strict: Wave 2 never retraces beyond Wave 1 start; Wave 3 is never the shortest impulse; Wave 4 never enters Wave 1's price territory. Fibonacci ratios define expected relationships: Wave 2 typically retraces 50–61.8% of Wave 1 (the golden pocket), Wave 3 extends to 1.618–2.618 × Wave 1, Wave 5 often equals Wave 1. These create high-probability entry zones that combine wave structure with mathematical precision.

Wave 3 entries (61.8% Wave 2 retracement confirmed): full 2% risk. Wave 5 entries: 1% only — Wave 5 often truncates and momentum diverges. Corrective wave trades (shorting Wave B): 0.5% risk — corrections are complex. Fibonacci extension targets: take 33% at 1.272× extension, 33% at 1.618×, trail remainder. Stop for Wave 3 entry: below the start of Wave 1 (invalidation rule).

Elliott Wave counting is notoriously subjective. The hard rules help but extended waves create ambiguity. Risk of misidentification is highest at market turning points — exactly where you most want to be positioned. Strict rule: if the next wave violates the count's rules, you have the wrong count — exit and re-analyse. Never argue with the market about the wave count. Extensions are the most dangerous: a third-of-third can look like a Wave 3 top prematurely.

Volume Profile maps where the most volume has traded over a given period, revealing price levels that matter most to market participants. The Point of Control (POC) — the price level with the highest traded volume — acts as a powerful magnet. Markets frequently gravitate back toward the POC because it represents where buyers and sellers reached the most agreement. The Value Area (70% of volume) defines fair value for the period. Breaking above VAH with expanding volume signals strong institutional demand. Falling below VAL signals aggressive selling.

POC-based trades (fading back to POC from outside Value Area): 0.75% risk — stop beyond the opposite VA boundary. VAH/VAL breakout trades: 1–1.5% risk — stop back inside the Value Area. Size up when multiple volume profile lookback periods agree on the same level. Scale out: take 50% at POC when approaching from outside, let 50% run to the opposite VA boundary.

Volume profile shows the past, not the future — levels that were significant can be taken out with minimal resistance if the context changes. Always use the most recent relevant timeframe. In thin markets (altcoins on LMEX), volume profiles are noisy and unreliable — stick to BTC-USDT and ETH-USDT. POC drift: as new volume prints, the POC slowly migrates. Your analysis can be valid at entry but invalidated an hour later as the profile updates.

Ichimoku Kinko Hyo ("equilibrium at a glance") is a complete trading system providing trend, momentum, and support/resistance analysis from a single indicator. The Cloud (Kumo) is the most important component — its colour and thickness define trend quality. The Tenkan/Kijun (TK) cross is the trigger. The Chikou (lagging span) confirms the signal by comparing current price to historical price 26 periods back. The most powerful entries occur when all elements align: price above a bullish Cloud, TK crossing above Kijun, and Chikou above historical price.

Scale size to signal quality. A+ signal (all 5 components aligned): 2% risk. B signal (price above cloud, TK cross but weak cloud or Chikou neutral): 1% risk. No trade inside the cloud — this is the noise zone. Stop placement for longs: below the bottom of the Cloud (Senkou Span B for thin clouds, Span A for thick). Trail stop with the bottom of the cloud. Target: measure the prior swing, project same distance from entry.

In fast-moving markets, Ichimoku lags significantly — signals confirm moves already substantially complete. Use lower timeframe TK crosses (1H) to fine-tune entries within 4H/Daily signal direction. The Cloud creates wide stops which forces smaller position sizes — intentional but reduces return on capital. Flat/ranging markets produce meaningless TK crosses near the cloud — enforce the no-trade-inside-cloud rule strictly.

Multi-timeframe confluence is a framework that dramatically improves the quality of any strategy. A signal on a 5m chart has very different predictive power depending on whether the 4H and daily charts agree. The principle is trend permission: the higher timeframe grants or denies permission for lower timeframe signals. A perfect RSI divergence on 15m is tradeable if the 4H is bullish; the same signal is ignored if the 4H is bearish. The weighted scoring system (Daily = 4×, 4H = 3×, 1H = 2×, 15m = 1×) quantifies confluence objectively.

Scale risk directly with the weighted confluence score. Score > 0.7 (strong alignment): full 2% risk. Score 0.5–0.7: 1% risk. Score 0.3–0.5: 0.5% risk or skip. Score < 0.3: no trade. For exits: close 50% when LTF signals reversal; close fully when MTF (1H) signals reversal; only let run to full target if HTF (4H+) remains aligned.

Conflicting timeframes are normal and expected — accept that a score of 0.6–0.7 is sufficient for a trade. Don't over-filter. Timeframe lag is a real risk — the daily signal may have triggered 2 days ago, making the LTF entry very late in the daily move. Check that the HTF signal is not exhausted (e.g., daily RSI at 85 when wanting to go long on 5m). Conflicting signals in adjacent timeframes often indicate a turning point — reduce size significantly.

The WTI/Brent spread exists because they are chemically similar crude oils traded at different benchmark hubs — Cushing, Oklahoma for WTI; North Sea for Brent. Historically, Brent trades at a $1–5 premium to WTI due to transportation economics and sulphur content. When the spread widens significantly beyond this norm (geopolitical supply shock, pipeline disruption, US export surge), it is expected to mean-revert. On LMEX, both trade as USDT-margined perpetuals (OIL-PERP and BRENT-PERP) — equal dollar positions on both legs, no physical delivery, no expiry roll. The ADF test confirms stationarity before any trade is placed.

Equal dollar value on both legs — non-negotiable for delta neutrality. Start at $500/leg, scale to $2,000/leg with track record. Leverage: 2–3× maximum — excessive leverage can be liquidated during transient volatility even when the thesis is correct. Set ISOLATED margin on each leg independently on LMEX. Target: Z-score returning from ±2.0 to ±0.5. Hard stop at Z = ±3.5.

Non-stationarity periods: OPEC+ production cuts or geopolitical embargoes can permanently shift the spread. Re-run ADF weekly — if p > 0.05, close immediately. EIA inventory data every Wednesday at 15:30 UTC causes sharp WTI-specific moves — avoid new positions Tuesday/Wednesday. Max hold time: 72 hours hard limit prevents riding a structural dislocation indefinitely. Funding on both perp legs accumulates as a carry cost.