A useful thing happened on this timeline this morning, and it is worth writing down.
An on-chain dashboard published a wallet as a smart trader: four trades, all bitcoin longs, a 100 percent win rate, 9.26 million in profit. Ninety minutes later the same dashboard published that the same wallet had been liquidated four times in fourteen hours, with 375.8 BTC of shorts wiped out. The winning record was long. The liquidations were short. Same address, opposite side.
The easy reading is the wrong one. This is not a story about one trader being bad. It is a story about what four observations can and cannot tell you.
Four trades winning is a one in sixteen outcome under a coin flip, about six percent. Dashboards track tens of thousands of active addresses, so a screen like that surfaces hundreds of perfect records by chance alone. And they get surfaced precisely because they are winning, which means the sample is chosen on the outcome you are trying to explain. The published record is a high-water mark by construction. That is why the gap between the celebration and the liquidation tends to be short.
The window matters as much as the record. Bitcoin ran from the high 70s to nearly 87,000 in a week. Four long trades across that stretch describe beta, not skill. Almost everyone who was long won.
Here is the part most commentary skips: the four liquidations do not establish the absence of skill either. A small sample says nothing in either direction. That was the whole point, and it cuts both ways. The honest sample for separating skill from variance is closer to fifty trades, across conditions that were not all the same.
If you are thinking about following an address, the numbers that would actually help are not win rate and not PNL. They are position size relative to the wallet's own balance, behaviour through the worst drawdown, and performance across a full cycle rather than one directional week.
What would you need to see from an address before you were willing to size behind it?
An on-chain dashboard published a wallet as a smart trader: four trades, all bitcoin longs, a 100 percent win rate, 9.26 million in profit. Ninety minutes later the same dashboard published that the same wallet had been liquidated four times in fourteen hours, with 375.8 BTC of shorts wiped out. The winning record was long. The liquidations were short. Same address, opposite side.
The easy reading is the wrong one. This is not a story about one trader being bad. It is a story about what four observations can and cannot tell you.
Four trades winning is a one in sixteen outcome under a coin flip, about six percent. Dashboards track tens of thousands of active addresses, so a screen like that surfaces hundreds of perfect records by chance alone. And they get surfaced precisely because they are winning, which means the sample is chosen on the outcome you are trying to explain. The published record is a high-water mark by construction. That is why the gap between the celebration and the liquidation tends to be short.
The window matters as much as the record. Bitcoin ran from the high 70s to nearly 87,000 in a week. Four long trades across that stretch describe beta, not skill. Almost everyone who was long won.
Here is the part most commentary skips: the four liquidations do not establish the absence of skill either. A small sample says nothing in either direction. That was the whole point, and it cuts both ways. The honest sample for separating skill from variance is closer to fifty trades, across conditions that were not all the same.
If you are thinking about following an address, the numbers that would actually help are not win rate and not PNL. They are position size relative to the wallet's own balance, behaviour through the worst drawdown, and performance across a full cycle rather than one directional week.
What would you need to see from an address before you were willing to size behind it?