Fog-of-war replay is market replay that deliberately hides future price action, forcing you to make trading decisions in real time as if the outcome hadn't happened yet. The point isn't a prettier backtest. It's building decision quality under genuine uncertainty. Use it when you're refining a strategy's entry logic, drilling execution discipline, or working on the behavioral habits that only show up when you don't know what happens next.
TL;DR:
- Fog-of-war replay trains decision-making in real time by hiding future price action, unlike traditional backtests that reveal outcomes immediately.
- Setting up the replay carefully across different market regimes and with realistic execution assumptions ensures meaningful, repeatable improvement.
- Avoid look-ahead bias by delaying fills and recalculating indicators using only available data, as leaks can artificially inflate performance metrics.
- Tracking specific KPIs—such as recognition speed, fill rate, slippage, and rule adherence—helps measure progress and focus on continuous improvement.
- Using dedicated platforms that integrate fog-of-war replay, journaling, and AI trade autopsies simplifies consistent practice and behavioral correction.
Table of Contents
- What Fog-of-War Replay Actually Is (and Why It's Not a Backtest)
- How to Set Up a Fog-of-War Replay Session
- Simulating Realistic Execution: Fills, Slippage, and Liquidity
- Common Evaluation Pitfalls and How to Catch Them
- How to Analyze Replay Sessions and Measure Decision Quality
- Five Practical Drills to Build Decision-Making Under Fog
- Author Tony's Perspective: What Fog-of-War Practice Actually Reveals
- Practicing Fog-of-War Replay Inside Discipline AI
- Sources
- FAQ
What Fog-of-War Replay Actually Is (and Why It's Not a Backtest)
A fog-of-war session takes historical price data and feeds it to you bar by bar, tick by tick, or however granularly you choose, without letting you peek ahead. You see a chart forming in October 2022 the way you would have seen it live, then you decide: enter, wait, size up, stand aside. No hindsight, no cheating.
That distinction from a standard backtest matters more than most traders assume. A backtest runs your rules against the whole dataset at once and reports a result. Live paper trading has real timing but takes weeks to accumulate enough reps. Fog-of-war replay compresses the timing pressure of live trading into minutes, minus the capital risk. It isolates the decision itself, stripped of the hindsight advantage that makes every backtest look smarter than the trader actually was in the moment.
Market replay is a well-established evaluation method, though research on strategy testing methods notes it doesn't make the market respond to your trades. That's fine for decision training. It's a limitation worth knowing before you draw conclusions about market impact.
How to Set Up a Fog-of-War Replay Session
A replay session only teaches you something useful if it's built to be repeated and compared, not just clicked through once. Sloppy setup is the single biggest reason traders run replay for months and never actually improve.
- Choose data windows deliberately. Don't just replay your favorite winning month. Pull segments across different regimes, a grinding chop, a trending breakout, a volatility spike, a liquidity crunch, so your skills generalize instead of overfitting to one market mood.
- Pick your timeframe, instrument, and account rules up front. Decide whether you're drilling five-minute crypto scalps or four-hour forex swings, and set your virtual account size, leverage, and margin rules to match what you'd actually trade with.
- Set your replay controls before you start, not mid-session. Fix your playback speed, note your session seed, and decide where you're allowed to pause or step forward. Simulators built for this let you replay historical data at controlled speed while genuinely hiding future bars.
- Write down your execution-model assumptions. Fill logic, slippage estimate, commission structure. Save all of it alongside the seed so the run is auditable later.
Pro Tip: Run the identical seed twice, a week apart, with the same rule set. If your decisions diverge significantly between the two runs, the problem isn't the market. It's inconsistent execution of your own plan.
Simulating Realistic Execution: Fills, Slippage, and Liquidity
A replay session that ignores execution mechanics teaches false confidence. If every fill happens instantly at your requested price, you're training against a market that doesn't exist.
Fill mechanics need order-book context, not just a last-traded price. A limit order in a thin order book might not fill at all; a market order during a liquidity gap might fill three ticks worse than the quote you saw.
Build your execution model around a few honest components:
- Apply a slippage estimate scaled to the instrument's typical spread and the volatility of the window you're replaying, not a flat number across every session.
- Charge realistic commissions and any funding or overnight costs that would apply live.
- Reserve full market-impact or agent-based modeling for strategies that push meaningful size relative to the instrument's depth. Plain replay tends to understate the permanent price impact of larger orders since the simulated market doesn't react to you.
- Accept the tradeoff: more execution realism slows down how many reps you can run per session. For pure decision-recognition drills, simpler fills are fine. For execution drills, spend the extra setup time.
Common Evaluation Pitfalls and How to Catch Them
Look-ahead bias is the quiet killer of replay training. It happens when your setup gives you information you couldn't have had at decision time, a same-bar fill that assumes you traded at a price before it was confirmed, or an indicator centered on future bars instead of computed causally as data arrives.
The damage is bigger than most traders expect. Controlled studies on backtest leakage show that a single surgical leak, like filling on the same bar a signal triggers, can manufacture Sharpe ratios that look dramatically better than the strategy's real edge.
Run these checks on any replay pipeline before you trust its results:
- Shift-fill test: delay every fill by one bar and see if your edge survives. If performance collapses, you had a look-ahead leak.
- Causal feature recompute: rebuild every indicator using only data available at that timestamp, then rerun the session.
- Rolling walk-forward: test on a window, then roll forward to fresh out-of-sample data rather than reusing the same period you tuned on.
- Overfitting red flags: a strategy that only works on one specific data window, or that needed a dozen parameter tweaks to look good, is memorizing noise. Document your data period and methodology honestly, since credible hypothetical performance disclosure depends on showing your work, not just your results.
How to Analyze Replay Sessions and Measure Decision Quality
A replay session without measurement is just entertainment. The value comes from logging specific numbers you can compare across sessions and watch trend in the right direction.
Track three categories:
- Decision-quality metrics: how many bars passed between a valid setup forming and you recognizing it, how often you followed your own predefined rules versus improvised, and what percentage of valid signals you actually acted on.
- Execution metrics: your fill rate on limit orders, average slippage per trade against your model's estimate, and the gap between the price you expected and the price you actually got.
- Behavioral metrics from your journal: stop-loss adherence, whether your position sizing stayed consistent with your plan, and how often stress or a losing streak triggered a rule break.
Journal every session as you go, not from memory afterward, and treat automated autopsy tools as a starting point for the "why," not the whole answer. When an AI trade autopsy flags a repeated pattern, like widening stops after a loss, that becomes your next drill, not a footnote.
Pro Tip: Pick one KPI to improve per week instead of tracking all of them at once. Trying to fix recognition speed, sizing discipline, and stop adherence simultaneously usually improves none of them.
Five Practical Drills to Build Decision-Making Under Fog
Generic replay ("just click through some charts") rarely sticks. Structured drills with a pass/fail bar do.
- Speed recognition drill: replay at 4x normal speed and score how many valid setups you correctly identify before the window closes.
- Stop discipline drill: commit to a stop before each trade and measure how many times across 20 reps you moved it after entry.
- Large-order handling drill: replay a session with a big resting order visible in the book and practice reading how price behaves around it.
- Regime-switch drill: start a session in a trend, let it flip into chop mid-replay, and track how fast you adjust your approach.
- Stress replay: compress three losing trades into rapid succession and log whether your sizing and stop behavior held up.
Set a numeric pass threshold for each, like 80% correct setup recognition, before moving to the next difficulty tier, and save every seed so you can rerun the exact same conditions later.
Author Tony's Perspective: What Fog-of-War Practice Actually Reveals
The traders who resist fog-of-war replay are usually the ones with the most to learn from it. Backtesting flatters you; replay doesn't. It exposes the half-second of hesitation, the stop you'd have moved, the size you'd have doubled out of impatience. Ad-hoc backtesting optimizes a curve. Disciplined, auditable replay drills change behavior, because you can't argue with a logged pattern that repeats across ten sessions. Build standing tests into your routine, not just standing rules.
— Tony
Practicing Fog-of-War Replay Inside Discipline AI
Everything covered above, session seeds, execution assumptions, KPI tracking, drill structure, is easier to run consistently when one platform handles setup and measurement together instead of juggling spreadsheets and separate tools.
Discipline AI's market replay simulation builds fog-of-war directly into the practice environment, hiding future bars while giving you speed and step controls, so you can run the drills above without building your own harness. The platform pairs that with automated trade journaling and performance analytics that log the decision and execution KPIs this article walks through, plus AI trade autopsies that flag the behavioral patterns worth drilling next. If you want a structured path through it, the Disciplined Trader program is a one-time $79 offering built around exactly this kind of disciplined practice. For ongoing use, Pro plans start at $8.99 per month, with annual and lifetime options listed on the same page. Browse the Learning Center for setup guides, or check current plans on the pricing page before your next replay session.
Sources
- How to Evaluate Trading Strategies: Single Agent Market Replay or Multiple Agent Interactive Simulation?
- How Much Sharpe Does a Look-Ahead Leak Manufacture? A Controlled Study of Three Surgical Leaks Against Known Ground Truth
- Replay Trading With Real Historical Market Data
FAQ
How Is Fog-of-War Replay Different From a Regular Backtest?
A backtest evaluates your rules against the full dataset at once, so you already know the outcome while designing it. Fog-of-war replay hides future bars and forces real-time decisions, which trains judgment rather than just measuring a rule set's historical fit.
How Do You Avoid Look-Ahead Bias in a Replay Setup?
Run a shift-fill test that delays fills by one bar and rebuild indicators so they only use data available at that timestamp. If your results change significantly after either fix, your original setup had a leak, as controlled leakage studies demonstrate.
Which Metrics Actually Matter When Reviewing a Replay Session?
Track recognition latency, rule adherence, and signal acceptance rate for decision quality, plus fill rate and slippage per trade for execution. Behavioral metrics like stop-loss adherence, logged through journaling, round out the picture.
Does Discipline AI Support Fog-of-War Replay?
Yes. Discipline AI's market replay feature hides future price data during sessions and pairs it with automated journaling and AI trade autopsies to track decision quality over time. Pro plans start at $8.99 per month, with annual and one-time purchase options also listed.
Should I Simulate Market Impact in Every Replay Session?
No. Plain replay works fine for most decision and execution drills, but strategies trading size large relative to an instrument's depth need agent-based or market-impact modeling, since standard replay can understate how much a large order would move the market.