Base Rates: The Question That Beats Pattern Recognition

A base rate is how often something happens across a reference class of similar situations. "What percent of breakout trades on this asset historically follow through?" "What percent of post-FOMC rallies sustain past 1 week?" "What percent of 5-trade winning streaks for me are followed by another winner?" Asking these questions before any trade decision is one of the highest-leverage cognitive corrections available, and one of the most universally underused.

What base rates are

Imagine you're trying to decide whether to take a specific trade. You can think about the trade in two ways:

Inside view (case-specific): "This breakout looks strong. Volume is good. Trend is up. I think it'll work."

Outside view (base rate): "Across all my breakout trades on this asset over the past 200 trades, the win rate is 45% with average winner +1.5R. That's the base rate."

The inside view focuses on what's special about this case. The outside view focuses on what typically happens in cases like this. Both are useful, but the outside view is the one most people skip.

The base-rate question is: given everything I know about historical situations like this one, what's the typical outcome? Then I can ask: is there anything genuinely different about this case that justifies updating from the base rate?

Why base rates beat pattern recognition

The brain is wired for case-by-case pattern recognition. "This setup looks like the one in March 2024 that ran 8R", the brain confidently extrapolates from a specific memorable example.

But specific memorable examples are systematically biased:

• You remember vivid winning trades more than slow losing ones
• You forget the trades that looked similar but failed
• You underestimate the variance across "similar" setups
• You over-weight recent examples

Base rates are the corrective. They aggregate across many cases, including the ones you don't remember vividly. "45% win rate" includes both the spectacular winners and the boring losers. It's a more accurate prior than the single example your brain pattern-matched on.

How to use base rates in trade decisions

Three steps:

1. Identify the reference class. What's the appropriate group of similar trades? "Breakout trades on BTC" is broad. "Breakout trades on BTC during trending regimes with above-average volume" is more specific. The narrower the reference class (within reason), the more relevant the base rate.

2. Look up or estimate the base rate. For your own trades, this comes from your journal. "Across the last 30 BTC breakout trades I've taken, the win rate is 50% with average winner +1.8R." For broader market base rates, requires either your own data or others' research.

3. Adjust if there's a specific reason. Is there something genuinely different about this case? "On-chain accumulation is unusually strong; I estimate +5% probability of follow-through." The adjustment should be small unless the difference is clearly unusual. Most of the time, the base rate is the best estimate.

The base rate is your prior. Specific case information adjusts the prior, slightly. Most retail goes the other way, treating specific case information as the primary input and ignoring the base rate entirely.

Base rates worth knowing for crypto

Some useful base rates for context:

• Breakout follow-through: roughly 40-55% of breakouts on liquid assets sustain past 24 hours, depending on regime. Most fakeouts.
• Trend reversal at major resistance: roughly 30-40% of resistance tests result in reversal; ~60% lead to break-and-continue.
• Capitulation low recovery: in major capitulations (sustained sentiment lows), recovery within 6 months has been ~70% historically, but the same period sometimes saw further declines first.
• Random altcoin recovering 100% from cycle drawdowns: most don't. ~80% of cycle losers never recover their prior highs.
• Short-streak continuation: a 3-trade winning streak has roughly base-rate probability of being followed by another winner, i.e., the streak doesn't predict the next outcome much.

These are rough, your specific reference class will have different numbers, but they illustrate the discipline of having quantitative priors instead of gut estimates.

A common mistake: ignoring base rates

A trader sees a beautiful breakout setup. They take it without considering the base rate of breakouts. The base rate of false breakouts on the asset in chop regimes is 60%; the trader didn't know that and didn't ask. They get stopped out and conclude "the setup looked perfect."

The fix: every "perfect setup" needs a base-rate sanity check. What % of "perfect setups" historically have worked? Almost always less than the trader's intuition suggests.

A common mistake: too narrow reference class

A trader looks for the base rate of "breakout setups on SOL during late-bull alt season with rising volume and on-chain confirmation in 2024." They find 2 historical examples; both worked.

Sample size of 2 is too small to give a reliable base rate. Either widen the reference class (lose specificity but gain sample size) or accept that you don't have a useful base rate for this exact situation.

The fix: aim for reference classes with 30+ historical instances. Broader classes with more data beat narrower classes with too few.

A common mistake: using overall market base rates for your trades

The "breakout success rate" published in some trading articles is across all traders, all setups, all regimes. Your specific base rate (your edge, your filtering, your execution) is different.

The fix: your journal is the most relevant source of base rates for your trading. The published numbers are useful for orientation; your numbers are useful for decisions.

A common mistake: treating the base rate as a prediction

The base rate says 45% of these setups work. That doesn't predict whether this specific one will. The base rate informs sizing and expectation, not the specific outcome.

The fix: base rate tells you what to expect across many trades. Each individual trade is still uncertain. The discipline is taking many positive-EV trades, not expecting any single one to deliver the average.

A common mistake: changing the base rate after each trade

After a losing trade, "the base rate must be lower than I thought." After a winner, "the base rate must be higher." Each trade triggers a recalibration that's not statistically warranted.

The fix: base rate updates require meaningful sample sizes. One trade barely moves a 30-trade base rate. Don't constantly recalibrate based on recent results; update when you have meaningful new data.

Bayesian updating, the correct way to adjust base rates

When you have a specific reason to think this case differs from the base rate, the correct adjustment is Bayesian. Without doing the formal math:

• Strong evidence: meaningful adjustment (e.g., 45% → 60%)
• Modest evidence: small adjustment (e.g., 45% → 50%)
• Weak evidence: negligible adjustment (e.g., 45% → 47%)

Most "this case is different" arguments are weak evidence. The base rate moves slightly, not dramatically. A trader who routinely adjusts the base rate by 20+ percentage points based on flimsy specific factors is overweighting the inside view.

A practical example

You're considering a long on a BTC breakout above $70k. You're confident, chart looks strong, on-chain supports.

Inside view: "This is a high-conviction setup. I think 70% chance it works."

Outside view (base rate): "BTC breakout setups in your journal have a 48% win rate over 60 trades. With +1.6R average winners and -1R losers, EV is +0.25R per trade."

Adjusted view: "On-chain confirmation is a real positive factor. Maybe 52% probability this specific one works. EV slightly higher than the base, ~+0.35R."

The adjusted view is more honest than the inside view's 70%. It's also actionable: positive EV, take the trade, size at 1% risk per the standard rules. Don't over-size based on inflated confidence.

This is what disciplined probabilistic trading looks like and it's significantly more accurate than gut-driven "this one's a sure thing" decisions.

Mental model, base rates as the gravitational pull on your estimate

Your case-specific intuition is like throwing a ball. The ball wants to fly in the direction your intuition points. Base rates are gravity, they pull the ball back toward the historical average. The ball still moves based on your throw, but gravity always brings it back toward the typical outcome.

A disciplined trader respects gravity. They throw the ball confidently when they have specific reasons but expect the ball's path to be mostly determined by gravity. An undisciplined trader ignores gravity and is constantly surprised when the ball doesn't fly where they pointed.

Why this matters for trading

Base rates correct most cognitive biases at once. They prevent overconfidence (your case isn't as special as you think), they prevent recency bias (recent examples don't change the base rate much), they prevent storytelling (the data-driven base beats the narrative intuition). Hex37's journal data is your primary source of personal base rates; over time, the rates by setup, regime, and asset become inputs to better decisions.

Takeaway

Base rates are the historical frequency of outcomes across a reference class of similar situations. Asking "what's the base rate?" before any trade decision corrects most cognitive biases. Identify the reference class, look up the rate, adjust slightly for genuine case differences. Most "this case is special" arguments are weak evidence, adjust the base rate slightly, not dramatically. The base rate is your prior; the specific case nudges it. Without base rates, you're trading on case-by-case impressions; with them, you're trading on calibrated expectations.