Down the Rabbit Hole: My Deep Dive into Crypto Prediction Markets Chapter 1: Discovered the Rabbit Hole

TheCodeMonkey — Fri, 07 Aug 2026 22:07:17 +0000

It all started on a not-so-good day. I didn't get much sleep the previous night, and I had a laboratory class to attend the next day. When I was about to leave for home, my friend called and said, 'I want you to work on the prediction market BTC-15M.' I didn't know what the prediction market was, so I did my research and was fascinated that something like this existed. Little did I know, this was the beginning of a really deep dive into a rabbit hole and it was not a good hole to have fallen into. Let me explain.

The first strategy I explored let us call it 'Sub-Candle Based Bias’.
I'm really bad when it comes to naming things, so yeah, deal with it :). The idea was to pick a timeframe for example, the 5M on BTC and check if the market ended up in a given direction matching that of the sub-candle. So if the 1M candle was bullish, then the 5M had a certain percentage of being bullish, and the same goes for the 2M, 3M, and 4M. This was the first idea I explored, and honestly, it was a straightforward idea simple to analyze and code.
After my analysis, I was able to get a win rate of 55–60%. This is not a completely accurate number, but it's close because I'm writing this from memory without looking at the analysis I did. After seeing this, I was not satisfied, so I added a filter, let's call it 'Candle Body Size.' This looks at the body size of the candle in relation to its total movement. This value was configurable: the less candle wick each sub-bar has, the more likely the market would close in that direction. So yes, this raised the win rate to about 76–80%, and I was happy. I mean, I thought I had just cracked the crypto prediction market and was going to be rich like, heyy yooo, this is money, right? Like, for real!! So yeah, I decided to do the actual test using Polymarket CLOB data. After building the sandbox, I ran some tests and found out my PnL was negative. I was shocked given that I had a 76–80% win rate, I should be on the plus side, right? But no. Turns out, in crypto prediction markets, if your strategy has an X% win rate, then your entry odds must be less than that X% for you to start making profit. Otherwise, you will lose. Let me explain.
In the prediction market, the higher the odds you buy at, the less profit you make. Buying at 90% odds gives you about 10% of your initial stake, and buying at 50% which is essentially a coin flip gives you 100% profit on your initial stake. So when you have a strategy with 80% accuracy and you're betting at 80% odds, you're only making 20% on your initial stake. Let me put this into perspective: say you entered 10 trades at $1 each and you won 8 of them using the 80% win rate, where you entered at 80%+ odds. Each win gives you $0.25 on the initial $1 stake, so in total you have $2. Remember, you lost 2 trades, which amounts to $2, leaving you with a PnL of $0.00. So yeah you just made breakeven you happy? SSo after I found out I tried doing something where I would only bet when the odds are below x% threshold and still did not work out, because my winrate reduced and caught up with the odds. So yes this strategy was now in the trash bag.

🚀 Physics Engine Progress Update I’ve made solid progress on my physics engine!

TheCodeMonkey — Mon, 29 Dec 2025 17:04:40 +0000

I have been building this ImpactPhysicsEngine for the past 3 months now and this is how far i've gone!! https://github.com/OtieteAyebanua/ImpactPhysicsEngine
there is a sample branch which shows how to implement the engine...

Hello everyone it's my first post here :)

TheCodeMonkey — Mon, 29 Dec 2025 14:28:33 +0000

DEV Community: TheCodeMonkey

Down the Rabbit Hole: My Deep Dive into Crypto Prediction Markets Chapter 1: Discovered the Rabbit Hole

🚀 Physics Engine Progress Update I’ve made solid progress on my physics engine!

Hello everyone it's my first post here :)