I bet civilisation ends by Christmas. Peace broke out.

Four deadlines, one settlement

Four of my Polymarket positions had deadlines in July, August, October and December. They settled within six minutes of each other, at 00:32, 00:37, 00:38 and 00:38 on 18 June.

The December contract settled six months before its deadline. I had no idea that was possible.

The six positions

Every one was a NO, which is a bet that the thing does not happen by the date on the ticket.

Lots are multiples of the smallest position.

Five of the six went to zero. That is 85.8% of the money in the cluster.

Why a December contract can die in June

I missed one sentence in the resolution rule:

This market will resolve to “Yes” if Iran and the United states agree to a permanent peace deal by the specified date, 11:59 PM ET. Otherwise, this market will resolve to “No”.

The date is a deadline, not a settlement date, and the rule does not need a signature. It resolves once both governments “provide clear public confirmation that a qualifying agreement has been definitively established.” The deal was announced on 14 and 15 June, and Polymarket settled all four contracts on 18 June, six months before the December one was due.

My stops had fired three days earlier, at 21:34 on 11 June, while a deal was still a rumour. I was not stopped out by the news. I was stopped out by everybody else working out what the news was going to be. The same asymmetry is why one position is still open: the YES price on Hormuz sits at 0.0015, and NO cannot be confirmed until the clock actually runs out. Being right pays late. Being wrong pays instantly.

The part that is worse than correlation

I had told myself the six positions were correlated. Four of them were not correlated. They were the same question asked four times.

Correlated positions move together. These were determined together, by one fact, at one instant.

No risk layer built on price history would have caught this. Correlation is measured on returns, and these contracts had barely enough history to measure. The relationship was never in the price series. It was in the resolution rules, which are plain English, free to read, and which nothing in my system was reading.

The correction I owe the previous episode

Episode 0 closed by promising to show why the overnight stop-loss cascade made the blowup worse. Now that the markets have settled, I can check, and the answer is that it made it better. Stopping out cost 34.5% of the cluster. Holding to settlement would have cost 80.9%. The rule I blamed for the damage is the only reason it was not much worse.

It fired in the wrong direction once. Hormuz was the one position I had right, and the same flat rule cut it at -30.1% on the way to a 34.4% win. A percentage stop cannot tell a wrong thesis from a right one under pressure. It charges you for both and you find out which was which months later.

The takeaways

1. Read the resolution rules before the chart. The rules decide what you actually own. Two contracts can look identical on the chart and still be the same bet. Only the rules tell you.

2. Nested markets are not diversification. If one outcome guarantees the other, holding both is one bet at double size. The same question at four different dates is the trap, because it looks like four bets and prices like four bets.

3. A stop-loss is not risk management. It only admits you cannot tell which position is wrong. Mine saved me because I was wrong five times out of six. That is luck, not a system.

4. A percentage stop behaves differently on every contract. Minus 30% fires after a small move on a cheap contract and needs more than twice as much on an expensive one. I set one number and got six different rules.

Where the book stands

90 closes, 60.0% win rate. Breakeven needs 60.61%, because the average loss runs 1.54 times the average win. The whole system sits about half a trade away from flat. The worst single close on the book is 13.2 times the average win, and it was another geopolitical bet.

What comes next

So I stopped trading geopolitics and started rebuilding for five-minute crypto binaries on the same venue. Next episode covers why, and it comes down to three properties this post has just described the absence of: a market that settles on a clock, a payoff that arrives at the same moment for both sides, and a config line that caps me at one open position.

Which kills the new setup first: fees, latency, or me?