Why is gold drawn to "prices nobody ordered"? — A story of a system that automatically trades gold using option OI

Have you ever found it strange when looking at the gold (XAUUSD) chart? Prices seem to stop at round numbers or at price ranges where there should be nothing, or conversely, they get pulled in like a magnet.
I believe this can be explained as a structure, not as "someone's presence." The key is CME gold option open interest (OI).
When a large amount of option open interest accumulates at a specific strike price, dealers on the opposite side perform delta hedging to neutralize their risk. Each time the price moves, hedge positions are mechanically bought or sold. This mechanical flow appears in the spot price as pressure toward a specific price range or as repelling pressure — this is not my own idea, but a phenomenon academically confirmed in stock markets as the "pinning effect." Combining this with the price impact framework of Kyle (1985), a classic in market microstructure, I am trying to systematize this for gold.
The system configuration is simple: it ingests CME option OI data daily, and a Python engine calculates a "pressure map" for each strike. The MT5 side monitors the speed of price changes, and only when both the map and speed conditions are met does a signal trigger and automatically enter a trade. Zero human discretion. Everything is logged.
By now, you must be thinking, "So, does it win?" To be honest, I don't know yet. I'm currently in the stage of verifying it.
And what I really want to convey starts here. In this field, graphs showing "90% win rate" and "30% monthly return" are everywhere. But an upward-sloping graph proves nothing. Anyone can win by going long in a bull market, and if you create 100 strategies, dozens will be positive by chance. In my own past verification, I once failed to detect the effect of my theory. When I investigated the cause, I found a flaw in the verification design — the data update frequency was insufficient, and I hadn't been able to isolate the conditions. It wasn't that the theory was wrong; it was that the verification wasn't designed to test the theory.
So now, I am first solidifying the verification protocol itself.
- Fix the hypothesis in advance. Do not start saying "Actually, I was testing this" after seeing the results.
- Do not draw a conclusion until at least 30 independent observations have been collected.
- Set a control group. Saying "it went up after condition A" alone says nothing. It only becomes meaningful when compared to times without condition A.
- Evaluate performance not based on "whether it is positive" but on "whether it exceeded the comparison target."
- Decide on falsification conditions in advance. Write down before verification which result would lead to abandoning the theory.
In short, the design is not to prove that it can win, but to quickly know if it is wrong. Only the theories that survive are worth betting money on.
Incidentally, this mindset is not limited to theory. In the early stages of system development, the signal performance deviated from the theoretical value, and for weeks I suspected a bug in the algorithm. The cause was not the code but the execution environment. Slippage and order delay were eating away the edge.
When I switched the execution venue to AfterPrime's MT5, the log numbers changed dramatically with the same code. If you want to use it, here you go.

Now, what I want to convey here is that the place to doubt is not intuition but what the logs tell you. The edge is only established when theory, implementation, and execution are all together. If any one of them is unverified, it is an unverified system.
If verification supports the hypothesis, I will move to the next stage. I intend to show the entire 100k yen challenge public trading from small capital on this account. Not just cherry-picking winning graphs, but including the design of the verification and even records of losses.
Details of the theory and verification results will be released here sequentially. If you are interested in market microstructure and option flows, please follow and see it through.