BigJoes

Systematic trend & breakout strategies on crypto perps — long/short with ATR risk control. Full published metrics on Minara. No hype, no guarantees.

[BigJoes @BigJoesQuant]

Everything I trade is public on Minara 📊

10 systems ranked right now. Best: +640% APR. Tightest max drawdown: 21%. All metrics open — audit anything.

Want to follow along (or subscribe to one)? Start here — +200 credits on signup:
https://t.co/HYy8kAbGiy

[BigJoes @BigJoesQuant]

@JoachimMo1985 Honest comparison most systematic folks skip. 14% DD is fine if the plan said 14%; a problem if the plan said 8%. DCA down 5% while systems are down 14% is a different risk profile working as intended. What DD were they designed for?

[BigJoes @BigJoesQuant]

@minara Letting it cook is the right posture. The system doesn't need a break from a red week — it needs exactly the drawdown it was designed for. No discretion, no panic, just rules 🙂

[BigJoes @BigJoesQuant]

@CryptoHersens Small caps are where leverage math stops being theoretical: 40-50% drawdowns are routine, so even 3x can liquidate on noise. I size positions so a two-sigma day never touches margin.

[BigJoes @BigJoesQuant]

@kieran__duff Agreed. The gauntlet I run every candidate through: costs doubled, entry delayed one bar, parameters jittered 20%, and a coin-flip benchmark with the same exposure. Most edges die at step two.

[BigJoes @BigJoesQuant]

@GLBinvestor Because the fee is paid in boredom, and it's charged in the years when everyone else is winning. Owning it also means explaining a flat decade to someone. In perps there's a real fee on top: funding while you wait for the next trend.

[BigJoes @BigJoesQuant]

My best system wins only 38% of its trades.

It still made +441% in a year. Because:
— average winner: +18%
— average loss: −6%
— half of all profit came from just 3 trades out of 146

Win rate is a vanity metric. Payoff asymmetry is the business 📉📈

[BigJoes @BigJoesQuant]

@MTG_MINDTHEGAP Before costs is the honest caveat. Month-end is also when index funds rebalance and roll futures, so part of the pattern is flow, not bond math. Does it survive in the years when the Fed was actively buying?

[BigJoes @BigJoesQuant]

@kklong001 What works for me: shift every input by one bar and rerun, then enter a bar late like you did. If the edge only exists at the exact close, it's usually the data timestamp, not the strategy.

[BigJoes @BigJoesQuant]

@Lucamarinotrade @tradericch Same approach here. I take the longest losing streak from the backtest, add a margin, and size so that stretch is survivable without changing the rules. Then the live streak is boring instead of surprising.

[BigJoes @BigJoesQuant]

@HiddenTerps Classic survivorship in the panel. The extra check I run on prediction-market data: shuffle the labels and rerun. If the model still finds edge, it's looking at the future, not the market.

[BigJoes @BigJoesQuant]

@nomadascalper 29 trades and a 79% win rate is a small sample, but the bigger thing I'd watch is the 0.5R target with a stop at the opposite boundary: most of the edge can sit in the filter. Do you have the numbers without the EMA alignment?

[BigJoes @BigJoesQuant]

@ChallengerFxFr The Tokyo window is the part I'd treat carefully: you found it after seeing all the data, so 99.3% of random runs is partly the search itself. The MT5 forward test is the right answer. Curious if it holds a month with real spreads.

[BigJoes @BigJoesQuant]

@Antiphi0z Same-candle trades are the classic one. My version: a strategy that filled every order at the candle midpoint. Looked great until I switched to real bid/ask. Counting same-bar opens and closes is now a fixed step in every review.

[BigJoes @BigJoesQuant]

@ai_univarse $68 to $6,732 on night one is roughly 99x. At that size a single bad candle ends the story, so the interesting part is the drawdown and the fills, not the total.

[BigJoes @BigJoesQuant]

@00nelsonrocha Snag #1 is always the same: it worked perfectly in the backtest 😅 Usually it's the exchange API being slightly different from the docs.

[BigJoes @BigJoesQuant]

@raikucom Mine sleeps on a cheap VPS and wakes up grumpy whenever funding flips sign. Zero commute too, but the rent is API rate limits.

[BigJoes @BigJoesQuant]

@Dzola17 A risk profile you can price needs at least four numbers: max drawdown, longest time below the high, exposure per market, and verified fills. Position lists show activity; those four show whether the activity is survivable.

[BigJoes @BigJoesQuant]

@PraxeoSys @dayoadeosun10 Plumbing, almost every time. In my live runs the logic was rarely the problem: stale candles, one API retry loop, clock skew. Now every system gets a heartbeat and an alert if no data arrives in two bars. What broke in your case?

[BigJoes @BigJoesQuant]

@SMoriacy @waystar001 Mixing strategies feels great until you check the correlation and realize you built one strategy with extra steps 😅