AArnold Holm3d ago · 4 min readCan a clean backtest still leave you with no trading edge?Can a clean backtest still leave you with no trading edge? If you check a report before you trust it, start with one ratio. Recorded profits divided by recorded losses tells you if gains kept more thastratcorealpha.hashnode.dev
AArnold Holm4d ago · 4 min readA weekly strategy still has a worst day: -9%Does a weekly basket rebuild keep you safe against a daily loss rule? If you run a challenge basket, that question matters more than the weekly plan. An illustrative $50,000 basket losing 9% ends at $stratcorealpha.hashnode.dev
SSergei B6d ago · 7 min readA Good Fill Price. What About the Fee?A limit order. You bought at exactly the price you specified, or even a little cheaper. So far, so good. But there is another line worth checking: what did the exchange charge for that fill? You can g1sergeialgo.hashnode.dev
SSergei B6d ago · 6 min readThe backtest used a 0.06% fee. What was my account actually paying?I develop trading bots. Our historical tests used a fee of 0.06% per side. When I checked the executions from one of my live accounts, though, the volume-weighted fee rate actually paid was about 0.0812Nsergeialgo.hashnode.dev
FForwardProofOct 25 · 7 min readWe tested 1,000 random trading rules on gold: how many look great?This article was first published on ForwardProof. This Hashnode post is a reprint; the canonical version lives on ForwardProof. Real hourly XAUUSD prices from 2015 to 2026, real bid/ask spreads, a cle11Aforwardproof.hashnode.dev
AArnold HolmOct 22 · 5 min read95% had no second funded account: keep four counters in your simulationIf you model reinvested payouts, count the steps before counting the next account. About 95% of the paths in our 24-month model never reached a second funded account. We simulated 20,000 paths startinstratcorealpha.hashnode.dev
AArnold HolmOct 19 · 4 min read$177 still open: does midnight erase your challenge loss?Your new trading day does not erase the loss on an open position. My MT5 simulation still carried a $177 open loss after midnight, with about $323 above its daily floor. I carry the open position acrostratcorealpha.hashnode.dev
LLaunch TapeOct 14 · 5 min readYour backtest is lying: 5 things to fix before you trust itYou wrote a bot, maybe with ChatGPT, Claude or Cursor doing most of the typing. You ran it over some history and it made money. That is the most dangerous moment in the whole project. A first backtest11Alaunchtape.hashnode.dev
AArnold HolmSep 28 · 3 min readQlib's US Region Does Not Make Commission Per ShareAn open Qlib question asks how to model US trading costs when the commission is charged per share. The example already sets trade_unit to one and passes a small number as open_cost. Those settings loostratcorealpha.hashnode.dev
GGlassboxtraderJul 14 · 3 min readEvery 'Pro' Risk Technique I Tested That Did NothingPart 7 of an honest build-in-public series — the wins, the dead ends, and what actually works. Previously: I Simulated 30x Leverage So You Don't Have To (It Went to Zero) There's an endless supply of1Aglassboxtrader.hashnode.dev
