Coding every PD array to determine its predictive value | Part 2: Fair Value Gaps. I coded fair value gaps on 5 years of Nasdaq data to see if they actually have any predictive value. 99% of them fill within minutes (which tells you nothing), tapping the gap barely beats random and doesn’t survive costs in any session, and the one entry that actually holds up isn’t what any guru teaches. In this series, we are going to code every one of ICT’s canonical PD arrays to determine its predictive value. It’s become too common and too widely accepted for internet, trading gurus to say that trading confluences “work” without showing any statistical proof of their predictive power. I hope this series will change that, and capture the attention of retail traders who need it. #statistics #education #quant #quantfinance
Full Transcript
and try to determine [music] if these are actually tradeable setups. The proximal gap tap breaks even at 1.96 ticks round turn and Nasdaq all in costs round [music] trip are usually two to four ticks, so that gets completely eaten by costs. You can't trade that. Testing all 24 hours, not just regular trading hours, that breaks even at 3.19 ticks, which is also probably below realistic costs. And then testing just overnight, that breaks even at 1.95 ticks, so that's also dead. Even the best hour, which is our 9:00 a.m. cash open one, breaks even around [music] three to four ticks. But if we do a much deeper entry instead of entering at the close side fair value gap, if we enter at the open of the displacement candle, then we get plus 0.13R per trade, and then we break even around 10 and 1/2 ticks. So even if we estimate pessimistic [music] costs and we take off four ticks, we still net 0.08R. If we match up the same events, the deeper entry in the gap beats the proximal edge of the gap tap by 0.2R. If you're someone who trades these ICT concepts and trades fair value gaps, that's pretty huge information. That being said, we're still just testing a feature, an individual feature. There's nothing built around this, no actual strategy. There's no position sizing or entry or exit timing or anything like that, but this is still really useful information for [music] you if you happen to trade fair value gaps. I hope that seeing a popular retail trading smart money concept confluence coded and actually statistically tested inspires you to maybe do [music] the same. And maybe to start exploring ideas beyond what's just taught on YouTube by these gurus who just want your money. There's an entire world of trading to be learned and discovered yourself when you stop listening to people who don't actually have your best interest at heart. You start actually building and testing your own stuff programmatically.
Original Description
Coding every PD array to determine its predictive value | Part 2: Fair Value Gaps. I coded fair value gaps on 5 years of Nasdaq data to see if they actually have any predictive value. 99% of them fill within minutes (which tells you nothing), tapping the gap barely beats random and doesn’t survive costs in any session, and the one entry that actually holds up isn’t what any guru teaches. In this series, we are going to code every one of ICT’s canonical PD arrays to determine its predictive value. It’s become too common and too widely accepted for internet, trading gurus to say that trading confluences “work” without showing any statistical proof of their predictive power. I hope this series will change that, and capture the attention of retail traders who need it. #statistics #education #quant #quantfinance