The condition

The Fed raised rates on 16 September. Gold rose the next day. My long gold spreads expired worthless.

The trade

I'd been building a view since August: debt levels, issuance shifting to the short end, interest expense compounding. My read was that we're closer to fiscal dominance than monetary dominance, and that the Fed had no room to raise rates.

I expressed it with GLD call spreads.

The FOMC raised rates unanimously. I was wrong about the decision. That is not why I lost money.

What I was actually holding

I thought I held a position on monetary policy. I held one thing: gold, a specific distance higher, by a specific date.

That's the whole position. Debit spreads don't pay for direction. They pay for a distance travelled inside a window. Gold moved my way in that window. It didn't move far enough.

I had never written that down.

Two things follow once you do.

You see how little work the story is doing. My thesis said the Fed would hold. It never said gold would be some percent higher by a particular Friday. There was a missing step between the claim and the payoff, and because I'd never stated the payoff condition, I never noticed the step was missing.

The market has already quoted your odds, and you paid them. The debit on a spread is the price of the condition. I didn't need to hunt for a probability elsewhere — I'd already bought one.

I did hunt, as it happens. After entering I saw futures implying the hike was near-certain, and decided the market was wrong without a specific argument for why. That's a separate mistake, and one I'll come back to. It isn't the one that cost me.

The second time

In April I wrote Long VIX — not what I thought. I'd been convinced an oil spike would take VIX to 60. Afterwards I counted how often VIX had actually closed that high. Rarely. I'd sized into a rare event without checking how rare.

Counting was the right instinct, but counting comes second. Until you've named the condition, you don't know what to count. In April the class wasn't "geopolitical crisis," it was a VIX level on a monthly close. In September it wasn't "Fed holds," it was a gold move of a given size in a given time.

Both times, the story I was telling and the condition I was paid on were different objects. Both times I noticed afterwards.

What I'm changing

One line in the log before every entry:

What, how far, by when.

The condition, not the thesis. Then attribute the P&L to that condition afterwards. If the thesis was right and the condition wasn't met, the thesis was decoration.

Still open

I haven't dropped the fiscal dominance view, but a unanimous hike under an administration demanding cuts is evidence against a version of it, and I owe myself a statement of what would make me abandon it. That's the next post.

And the question I walked past: when a market prices something as near-certain, how often is it right? I don't know yet. That one needs data.