@tradingstrategies.workV10 — Strategy by market phase
The same rule earns for two years and bleeds for one, without changing a line. What changed is the market around it. One card splits a strategy's history by phase, and marks which phase we are in now.
@tradingstrategies.workV9 — Measure a filter's effect
Can a filter turn a mediocre rule into a good one? Every filter promises it. The Edge Library measures it: baseline vs. filtered, sample size, verdict, filter by filter. Promises get numbers.
@tradingstrategies.work10/10
Check what such an assistant says about itself against the interface. For market data the reverse holds: the answer is as good as the tool behind it.
tradingstrategies.work/blog/trading...
Study the Past — Improve your Future 🥋 #backtesting
@tradingstrategies.workYour rule works on the daily. Does it work on the 4-hour?
→ Elite opens every interval, not just daily
→ same strategy, same pair, different clock
→ fees included, so the fast one has to earn it
Free for a month if you register by end of October.
@benjamin-cup.bsky.socialYour backtest might be cheating. 👀
Look-ahead bias can make a strategy look profitable by using future data.
I break down the most common mistakes and how to avoid them:
@tradingstrategies.work6/6
Next large book: the 30 Oct monthly, 115,222 contracts, max pain 73,000. That strike sits 14 % below spot, with 70,050 calls above spot.
Watch whether it climbs too.
tradingstrategies.work/dashboard/bi...
Study the Past — Improve your Future 🥋
@tradingstrategies.workV7 — Check parameter robustness
The best setting you found: real skill, or the luckiest number? The robustness field puts your result next to its neighbours: same rule on other pairs, then the settings next door. Plateau beats peak.
@beefedai.bsky.socialIndependent Model Validation: Methods, Tests, and a Practical Checklist
https://beefed.ai/en/independent-model-validation-checklist
#ModelValidation #IndependentValidation #Backtesting #StressTesting #SensitivityAnalysis
@tradingstrategies.workThat backtest looks great. Is it real, or one lucky cell?
→ the robustness field runs the rule across neighbouring settings
→ a real edge holds up next door, a lucky one does not
Elite, free for 30 days, if you register by end of October.
@tradingstrategies.work10/10
Shown a multi-day-candle result? Ask:
→ which alignment
→ which window
→ how many trades
Under ~12 pp CAGR, 1d/2d/3d are one block, not a ranking.
tradingstrategies.work/dashboard/cr...
Not investment advice, not a forecast.
Study the Past — Improve your Future 🥋
#backtesting
@tradingstrategies.work1/10
The parameter you never set.
If your indicator runs on 2-day or 3-day candles, you never chose which calendar day the candle starts on. The grid is anchored to the Unix epoch.
That choice alone moves CAGR by 6.66 pp on average.
@tradingstrategies.workV6 — Understanding the Arena Score
Three checks before a result deserves trust: benchmark, costs, sample size. The Arena Score weighs them into one number and shows its parts. How to read it, and what it cannot tell you.
@tradingstrategies.workDCA or lump sum? That is a question about history, not opinion.
→ run both on Bitcoin back to 2010
→ same window, same fees, side by side
→ see the drawdown you would have sat through
Free Elite month for accounts registered by end of October.
@tradingstrategies.work10/10
Before a backtest moves money:
→ universe first: BTC only, or alts plus strategy
→ regime gate before indicator
→ alts against BTC
→ neighbourhood, fees, drawdown
tradingstrategies.work/blog/strateg...
Study the Past — Improve your Future 🥋
@tradingstrategies.work1/10
You tested a strategy on a coin. It lost 10 % a year. So it's no good?
Read the line under it: what the coin did with no strategy at all. Across 197,435 backtests on 653 crypto pairs, that second line is the whole story.
@tradingstrategies.workV5 — Trade count and sample size
Flip a coin ten times and seven heads is not unusual. A strategy with a handful of trades works the same way. Check the trade count before you trust a backtest: where it hides, when the platform warns.
@tradingstrategies.workV4 — Reading the result
Total return looks great on screen. Which part survives costs, and which year carried the whole gain? The result panel line by line: the figures, the year table, net of costs.
@tradegrub.bsky.socialMost traders stack 5+ indicators and wonder why backtests crush live. The real problem: each indicator adds noise unless it answers a specific question. Build layers that filter, not confirm. Test each one independently first. Remove what doesn't reduce drawdown. #trading #backtesting #indicators
@tradegrub.bsky.socialWalk-forward optimization isn't magic. You train on past data, test on future data, then shift the window forward and repeat. Most traders treat it like gospel. #Backtesting #AlgoTrading #TradingStrategy
@tradingstrategies.work9/9
After Friday 185k contracts roll off. Next:
→ 30 Oct monthly 101,103, max pain 73,000
→ 25 Dec quarterly 118,043, max pain 75,000
Watch whether the strike chases spot.
tradingstrategies.work/dashboard/bi...
Study the Past — Improve your Future 🥋 #backtesting