Table of Contents
Kevin Davey—champion algorithmic trader, contest winner, and author—sits down to unpack how he actually builds and runs rule-based systems across futures, forex, and more. He’s known for turning rigorous testing into practical, tradable ideas and for teaching everyday traders how to stop guessing and start validating. If you’ve ever felt stuck for strategy ideas or overwhelmed by coding, Kevin’s engineering-style approach will feel like a breath of fresh air.
In this piece, you’ll learn Kevin’s playbook for sourcing ideas from public places (books, forums, mags), pressure-testing them fast, and only scaling what survives costs, slippage, and Monte Carlo reality checks. We’ll cover why “good enough” backtests beat curve-fitted perfection, how to mix daily and intraday systems, and simple ways to diversify by market, timeframe, and logic to tame correlation shocks. You’ll also see his practical twist on entries and exits—treating many entries as exits to expand your toolbox without complexity—so you can turn research into a real trading workflow.
Kevin Davey Playbook & Strategy: How He Actually Trades
Idea sourcing: turn common market edges into testable systems
Great edges don’t need to be secret; they need to be testable. Kevin looks for simple, explainable patterns that have a logical driver—then turns them into rules he can verify fast.
- Start every concept with a one-sentence edge hypothesis (e.g., “Strong up days tend to follow a pullback in trending markets”).
- Pull ideas from widely known patterns (breakouts, mean reversion, seasonality, opening range, volatility crush) and write them as IF–THEN rules.
- Limit each system to 3–7 core rules (entry, exit, filter); if you need more, split it into separate systems.
- Use only daily or liquid intraday data at first; add complexity later only if the edge survives.
- Require a plausible economic/behavioral rationale for each rule before testing.
Research pipeline: from sketch to tradable in weeks, not months
Kevin runs ideas through a strict pipeline so only robust systems reach live capital. This prevents curve-fitting and keeps the portfolio fresh without gambling on unproven edges.
- Phase 1 (rough backtest): 2000+ bars, gross edge must exceed estimated costs by ≥2×.
- Phase 2 (parameter sweep): search wide but coarse; prefer flat plateaus over sharp peaks.
- Phase 3 (walk-forward/OOS): hold out at least 20–30% of history as true out-of-sample.
- Phase 4 (incubation/paper): trade signals in real time, 4–12 weeks; no code edits allowed.
- Phase 5 (small live): risk ≤0.25% per trade until live metrics match incubation. Promote only if the slippage and error rate are within the plan.
Data hygiene & backtesting standards: kill bad results early
Most “great” backtests die from dirty data or unrealistic assumptions. This step keeps you honest, so the performance you see is the performance you can actually capture.
- Use survivorship-bias-free data where applicable; for futures, stitch continuous contracts with back-adjustment.
- Model realistic costs: commission, exchange fees, and slippage (e.g., 1–2 ticks per side for liquid futures; adjust by volatility).
- Enforce trade-through rules: entries/exits only at available prices; no look-ahead on highs/lows of the same bar.
- Cap daily/weekly trade count to what you can execute; reject systems that rely on micro-scalps in illiquid hours.
- Require at least 100 independent trades in test+OOS before considering live.
Walk-forward validation: prove adaptability, not perfection
Markets change; robust systems adapt without frequent retuning. Kevin prioritizes walk-forward methods to test stability over time.
- Split history into rolling windows (e.g., 3–12 months optimize, next 1–3 months trade); compute walk-forward efficiency ≥50%.
- Prefer parameters that remain inside a broad “good” zone across windows.
- If the best parameter changes wildly each window, treat it as a curve-fit and reject.
- Track OOS Sharpe/Sortino within 70–110% of in-sample targets; outside that, re-evaluate logic.
- Keep a walk-forward schedule on a calendar; re-optimize no more than quarterly unless volatility regime shifts dramatically.
Robustness testing: try to break it before the market does
A system that only works under perfect conditions isn’t a system. Stress tests simulate messier realities to see if the edge survives.
- Monte Carlo resample trade sequences 1000+ times; require that the 5th-percentile equity path is still acceptable.
- Shock slippage/commission by 50–200% and reduce fill rate by 10–20%; system must remain profitable.
- Randomize entry/exit prices by ±0–1 tick; if PnL collapses, the edge is too fragile.
- Delay entries by 1 bar and exits by 1 bar; require profitability to persist.
- For time-of-day systems, shift session boundaries by ±5–15 minutes to test sensitivity.
Entries & exits: simple triggers, disciplined exits
Kevin treats entries as hypotheses and exits as risk control. Keep the trigger simple and the exit rules mechanical to avoid hesitation.
- Use one primary trigger (breakout, pullback, momentum cross) plus at most one filter (trend or volatility).
- Always include a hard initial stop based on ATR or structure (e.g., 1.5–2.5× ATR or below last swing).
- Add a catastrophic safety stop (e.g., exchange circuit levels, max loss per day).
- Employ time-based exits to cut dead money (e.g., exit after N bars if TP/SL not hit).
- Consider partial profit at 1R and trail the rest with a volatility-adjusted stop (e.g., 1× ATR below highest close).
Position sizing & risk: survive first, compound second
He focuses on staying power: modest risk per trade, portfolio-level limits, and rules that reduce exposure during drawdowns.
- Risk 0.25–0.75% of equity per position; cap portfolio at 2–3% total risk at any time.
- Size by volatility: contracts = (risk_per_trade) / (ATR_k * dollar_per_point).
- Use correlation-aware caps: if two systems trade the same market/direction, cut size by 50% each.
- Implement an equity curve stop: if peak-to-trough reaches 1.5–2× your tested max DD, reduce size by 50% or pause.
- Rebalance monthly; never scale up mid-drawdown.
Portfolio construction: uncorrelated systems across markets and logic
Rather than rely on one “holy grail,” Kevin stacks multiple modest edges. The goal is smoothness through diversification by market, timeframe, and setup logic.
- Target 8–20 active systems across equities, rates, energies, metals, ags, FX.
- Mix logic families: breakout, mean reversion, trend-follow, seasonality, volatility.
- Blend timeframes (daily plus intraday, like 15–60m) to reduce correlation spikes.
- Enforce market exposure limits (e.g., max 40% notional in any single sector).
- Replace underperforming systems only when a vetted newcomer improves portfolio MAR or reduces drawdown.
Execution & automation: reduce human error to near zero
Cleaner execution beats clever ideas. Kevin automates signal generation and uses tight operational checklists so live trading matches the backtest.
- Generate signals before the session; push orders with protective stops/targets attached (OCO).
- Use session-aware scheduling; avoid thin periods if slippage dominates your edge.
- Log every fill: timestamp, latency, slippage vs. model; investigate deviations >1 tick.
- If live slippage exceeds the model by 50% over 20 trades, cut size or pause that system.
- Keep a runbook for outages: broker down, data feed lag, and emergency flat-all procedure.
Monitoring & metrics: manage by numbers, not feelings
You can’t improve what you don’t measure. Kevin tracks a small set of stats that link directly to survival and compounding.
- Weekly: equity, drawdown, trade count, win rate, average R, profit factor, expectancy.
- Monthly: Sharpe/Sortino, MAR (CAGR/DD), tail metrics (worst 5 trades), heat map by market/logic.
- Flag if win rate or average R drifts >20% from test values for 50+ trades.
- Maintain a system “health score” (0–100) combining slippage, error rate, and tracking error vs. model; below 60 triggers review.
- Archive versioned code/configs so you can always recreate any month’s results.
Maintenance & retirement: keep the stable fresh
Every edge decays. Kevin plans for system retirement the same way he plans entries—by rule—so the portfolio stays robust over time.
- Define kill criteria up front: e.g., PF < 1.05 over 200 trades, or new max DD > 1.5× tested.
- If a system hits soft-warning thresholds, move it to incubation and trade at half size.
- Add two new vetted systems for every one retired to maintain diversity.
- Re-test annually with updated data; logic changes require a full pipeline rerun.
- Keep a backlog of 10–30 incubating ideas so replacements are ready when needed.
Trader mindset & workflow: process beats prediction
This approach removes guesswork and emotions from day-to-day decisions. The edge is in the process: consistent testing, strict risk, and steady iteration.
- Set a weekly research quota (e.g., 3 ideas screened, 1 progressed to parameter sweep).
- Time-box discretionary tweaks: if you can’t justify a change in one paragraph, don’t ship it.
- Treat live mistakes like bugs: root-cause, patch, and document in the runbook.
- Protect focus with a pre-market checklist and a post-market debrief (both under 10 minutes).
- Celebrate process compliance, not P&L; the former creates the latter.
Size every trade by volatility; cap portfolio heat to survive drawdowns.
Kevin Davey is ruthless about sizing because survival beats bravado. He starts with risk per trade as a small, fixed percent, then lets volatility set the contract count so each position has similar pain potential. That keeps a wild market from hijacking your P&L just because it moves more per tick. If ATR expands, size shrinks; if ATR contracts, size can rise—always within a strict max risk.
He also caps “portfolio heat,” the total open risk across positions, so a string of correlated trades can’t sink the account. When several systems point the same way in the same market, he cuts size or skips signals to prevent stack-on losses. Drawdowns get automatic brakes: reduce size when equity slips beyond a threshold, and only scale back up after recovery. With Kevin Davey’s approach, sizing isn’t a guess; it’s a rule that protects the account first and compounds only when conditions justify it.
Build simple rule-based systems; avoid prediction and reward mechanical execution.
Kevin Davey strips strategy design down to clear IF–THEN rules and lets the stats do the judging. No forecasts, no narratives—just triggers, filters, and exits that can be tested and repeated. He favors one clean entry condition and a small number of validations over kitchen-sink logic. The goal is consistency: the same decision, every time, without negotiating with the chart.
Mechanical execution is how those rules actually turn into returns. Kevin Davey predefines stops, targets, and timeouts so discretion can’t creep in mid-trade. He tracks whether live results match the model and pauses fast if slippage or errors exceed limits. When the rules fire, he follows them; when they don’t, he does nothing—and that discipline is the real edge.
Diversify by market, timeframe, and logic to smooth equity curves.
Kevin Davey spreads risk across uncorrelated engines so one bad day doesn’t define the month. He mixes markets—equities, energies, metals, rates, ags, and FX—so a shock in one sector isn’t a shock to all. Timeframes get the same treatment: daily trend systems pair with intraday mean-reversion or breakout models to balance behavior. The result is fewer synchronized losses and a steadier climb.
Diversification also means different logic families, so you’re not betting on one idea of how markets “should” move. Kevin Davey runs trend-follow, mean reversion, seasonality, and volatility plays side by side, each with its own rules and position caps. If two systems hit the same symbol in the same direction, he scales them down to avoid hidden correlation. This way, portfolio value depends on a team of modest edges—not a single hero trade.
Validate edges with out-of-sample tests, Monte Carlo, and stress scenarios.
Kevin Davey treats validation like a crash test for trading ideas. After the initial backtest, he holds back a clean slice of data for true out-of-sample and expects results to be in the same ballpark—no excuses, no curve-fitting. If the edge only shines in-sample, it goes in the bin. He also runs walk-forward checks to prove the system can adapt without constant retuning. The goal is a rule set that performs “good enough” across regimes, not “perfect” in one narrow window.
Then comes the abuse phase. Kevin Davey scrambles trade order with Monte Carlo to estimate how ugly the path can get and whether the 5th-percentile outcome is still survivable. He inflates slippage, delays entries, and jitters prices by a tick to see if the logic breaks. If profits vanish under small frictions, the edge wasn’t real. Only systems that pass these beatdowns earn live capital.
Define exits first; use stops, timeouts, and trailing rules for control.
Kevin Davey designs exits before he worries about clever entries, because control of risk is what keeps traders in the game. He uses a hard initial stop based on structure or AT, R, so every position has a known worst-case from the start. Time-based exits kick out dead trades that neither hit the target nor stop, freeing capital for fresher signals. Targets are pragmatic—often at a multiple of risk—so the system doesn’t depend on perfect trends.
From there, Kevin Davey lets trailing logic manage the winners without micromanagement. A volatility-adjusted trail moves only after meaningful progress, avoiding whipsaws while still protecting open profit. He scales partials at the first milestone and lets the remainder ride, giving the portfolio both consistency and home-run potential. When live slippage or tracking error deviates from the model beyond preset bounds, he cuts size or pauses—because exit rules aren’t suggestions; they’re the backbone of his process.
Kevin Davey’s core lesson is to build a portfolio of rule-based systems that diversify by market, timeframe, and underlying logic while never forgetting that correlations can spike when you least expect them. He spreads risk across futures sectors and mixes daily with intraday bars, then checks correlation and keeps an open mind about what actually delivers the best risk-adjusted returns. To prepare for the worst, he runs Monte Carlo to estimate how ugly drawdowns can get and accepts that even a basket of many systems can still line up against you during unusual events.
He also favors practical system construction: daily-bar swing systems are often easier to develop and still catch big trends, while intraday systems can be harder, even though they flatten by day’s end. On exits, he flips the script—treats many “entry” conditions as exits to react to price instead of relying on arbitrary dollar stops and targets, instantly expanding the exit toolbox. For evaluation, he leans on a simple benchmark: aim for a return-to-drawdown of roughly two to one, and set explicit goals that pair desired returns with a drawdown you can stomach. This is the heartbeat of his process: robust validation, smart exits, realistic goals, and diversification that respects correlation risk.
Related posts:
- Kevin Davey Trader Strategy: How an Algo Pro Really Builds & Trades Systems
- Larry Williams Trader Strategy: How the 11,000% Champion Really Trades
- Trader Strategy Playbook: How a World Cup Futures Trader Finds Edge Over Wall Street
- Trader Strategy Spotlight: Trader Nick on Building Real, Steady Edges
- Andrew Aziz Trader Strategy: How a Full-Time Scalper Finds High-Probability Trades
- Trader Strategy: Discipline, Timing, and Edges from a Veteran FX Pro
- From Nightlife Tycoon to Trader Strategy: Momo’s Playbook for Edges
- Brad Jelinek Trader Strategy: Flexible Edges for Manic Markets
- Martyn Tinsley Trader Strategy: Building Algorithmic Edges and Portfolio Discipline
- Brent Penfold Trader Strategy: How a Risk-First Mindset Builds Durable Edges
- Kevin Timmer’s Trader Strategy: Simple Levels, Real Discipline
- Trader Psychology Strategy: Brett Steenbarger on Building Edges That Last
