Oil Was Oil. Until It Wasn't.

Oil was oil. That's what you assume when you start backtesting. The ticker says WTI, the chart says oil, the broker says oil, your data feed says oil. Four different sources, all nodding at the same underlying thing. You move on.

That assumption survived for months in our research. It survived parameter sweeps, an exit strategy rewrite, several rounds of "wait, is this overfitted," and the optimistic "let's go paper" decision. It died on a Sunday afternoon when somebody pulled up the broker terminal and the data feed side by side and noticed that one screen said $93 and the other said $77.

Sixteen dollars apart. Same instrument, allegedly.

The first reaction was that something was broken. Stale cache. Wrong symbol. A config number you fat-fingered three weeks ago. You walk through the obvious things first because the obvious things are usually right. None of them were the explanation here.

The actual reason turned out to be more boring and more uncomfortable at the same time. The data feed was tracking the continuous front-month futures contract. The broker was offering a back-adjusted CFD, rolled and smoothed across contract changes. Both products call themselves WTI. Both move with oil prices, mostly. But they aren't the same number. They never were going to be.

Sixteen dollars of static gap was the headline. The real problem was underneath: when you look at twelve-hour windows and ask whether the two prices moved in the same direction, the answer was something between "maybe" and "barely." Not ninety-five percent. Not ninety. Closer to flipping a coin and being mildly disappointed.

That is not what you want to find out after you've already decided your strategy works.

So we did the boring sane response. Stopped, documented it, didn't blow up anything. The same idea got migrated to a broker whose oil contract actually does match the data feed it was researched against — prices line up, moves agree, the world makes sense again. The original broker stays in the system as an odd second observation point, useful precisely because it doesn't agree.

The lesson sounds obvious in writing and is genuinely easy to miss when you're heads-down in a backtest. "Same ticker" doesn't mean "same number." Two brokers can both call something WTI Oil and not be telling you the same story. If your strategy was researched on series A and traded on series B, you are running an experiment you didn't sign up for.

We knew about contract specifications. We knew about CFDs. We had read about back-adjustment. Knowing something abstractly and checking that it applies to the specific pipe your money is going to flow through are two different acts. The second one is the one that saves you.

Anyway. Oil is now agreeing with itself again. The book is smaller than we'd like, and the new numbers are going to need a few more weeks before any of them mean anything. But at least the thing we're testing is the thing we thought we were testing.

This is not investment advice. Past results don't guarantee future performance. All analysis reflects our internal research process only.