💡 In one sentence
"I think it'll bounce here" isn't a call — it's an emotion. Whether a call can be falsified is the one and only line that decides whether it has any value. This article teaches you to break gut feel into rules a machine can execute.
In "Seeing Through Hindsight Line-Drawing" we said: what truly separates signal from noise is a rule that existed beforehand and can be falsified. This article takes that "rule that existed beforehand" all the way down to something you can actually implement.
1. What does "falsifiable" mean?
A call is falsifiable if and only if: it states explicitly "under what conditions I am wrong."
❌ Not falsifiable
"Bitcoin is bullish long term." If it rises, it was right; if it falls, "the long term just hasn't arrived yet" — it can never be wrong, so it carries no information.
✅ Falsifiable
"If the daily closes below 58,000, my call on this rally is void and I exit." It states a specific condition that could prove it wrong.
Note: falsifiable ≠ frequently wrong. It means you have given a specific condition that could prove you wrong. A call that can never be falsified can never be verified as right either — it's just talk. This is the exact opposite of "hindsight line-drawing": the line-drawer never states an invalidation condition, because without one they can never be "wrong."
2. The 5 parts of a complete rule
Translating gut feel into a rule means breaking it into these 5 parts, all of which must be pinned down. Miss one and it's still just a "feeling."
By this point you'll notice: a decent rule is itself a piece of pseudocode. That's no coincidence — the end of falsifiable is executable.
3. From gut feel to rule: a practical workflow
- 1. Capture the observation: reduce "I think it'll go up" to what you actually saw — a volume spike? a long lower wick? a price level tested repeatedly? Write down that observable trigger.
- 2. Define it: quantify the fuzzy words. "Volume spike" → "volume > 1.5× the 20-bar average volume"; "long lower wick" → "lower wick ≥ 2× the body." Fuzzy words are the enemy of a rule.
- 3. Complete all 5 parts: add the stop, exit, sizing, and filter.
- 4. Write the invalidation condition: ask one more time on its own — "what would make me admit this rule has stopped working in the current market?"
- 5. Throw it into a backtest: run it on historical data and look at win rate, profit factor, max drawdown, and sample size. (How to read these numbers without fooling yourself: backtest > perfect trades)
4. The biggest trap: overfitting (curve-fitting)
Once the rule is written clearly, the trap beginners fall into most easily is endlessly adding parameters and tweaking values to make the backtest curve prettier, until it's nearly perfect on historical data. This is "hindsight line-drawing" coming back in a new outfit: you're using a known future (history) to reverse-carve a rule that only works on the past.
Warning signs
- Oddly "precise" parameters (a 2.7% stop, a 23-day MA) — usually the fingerprint of fitting noise.
- The rule collapses the moment you change anything — a robust rule isn't sensitive to small parameter tweaks.
- Too small a sample — a "70% win rate" over 20 trades has no statistical meaning.
The antidote: hold back out-of-sample data for testing, round parameters to robust ranges rather than the optimal point, and always remember — a backtest is only the entry ticket; live trading is the courtroom (live ≠ backtest).
5. The rule is written — who executes it?
Suppose you now have a rule that is falsifiable, passed its backtest, and held back out-of-sample data. Over the next 24/7, who executes it without emotion? People get tired, get greedy, hesitate when they should cut, and make exceptions because "this time feels different." The biggest enemy of a good rule is the person executing it. This is exactly why systematic execution exists.
In essence, CoinTech2u (the AI dynamic multi-strategy trading system) is exactly such a collection of rules:
Multi-strategy = multiple falsifiable rules
Each has explicit entry / stop / exit / filter conditions, has passed both backtesting and live validation, and dynamically shifts weight by market regime — rather than one "do-everything holy grail" forced through every condition.
Execution: zero emotion, zero exceptions
Once a rule goes live, the system executes it strictly — no shaky hands because "this time feels different."
Results you can falsify yourself
/live-proof reads the official system API directly every hour and lays out the real operating results — you can turn this article's standard right back on us.
Non-custodial
The strategy runs, but your money always stays in your own exchange account, with minimized API permissions and no withdrawal rights. Handing the rules to a system is the end of discipline, not its starting point.
Make yourself provable-wrong first, and only then can you be verified right
See what a set of falsifiable rules looks like running live — take this article's standard and judge CoinTech2u for yourself.