The worst day
We took one of the worst days crypto has had and ran 43,200 test Arlos through it, one second at a time. Here is what happened to the Backers’ money, and how we tested it.
12 March 2020
Bitcoin opened the day near $7,934. At 10:47 it fell 15% in under 5 minutes. By midnight it was 40% down on the day.
The test
Each test Arlo opened at a different moment. One opened every minute of the day, at six levels of leverage from 2x to 100x. Their stops sat at different distances. Each put all of its loss room to work in Bitcoin, Ether and Solana, up to the caps. Then we played the day’s prices through Arlo’s rules, second by second.
Stops filled a few seconds late, the way GMX fills them. The line closed every trade five seconds after the wallet reached it. Arlo paid the network cost of the trades closed at the end, as it does for real.
What happened
Almost every Arlo that fell short did so in the same way: prices fell so fast that trades closed well past their stops. The rest had a stop that did not fire. At low leverage the loss room covered it. At the highest leverage, each trade held so little margin that losing all of it still fit in the room.
Four rules did the work
- The Trader’s money goes first. Every loss comes out of the Trader’s $1,000 before any of the Backer’s $2,000.
- Every trade’s worst case must fit. The loss at its stop, or its whole margin with no stop, must fit in the $800 of loss room before it opens.
- Every trade has a cap. At most $1,000 of margin and $15,000 of size.
- The whole wallet is watched. If its value falls to $2,200, counting open trades, every trade closes.
What this test leaves out
- It is one day. A different day could go worse.
- Each test Arlo held three long trades. A Trader who spreads small, high-leverage trades across many markets just before a crash can lose more. In the worst case that takes the Backer’s whole contribution. The Trader’s money still goes first.
- It leaves out how big orders move prices. On a real day, closing many large trades at once can cost more.
- It is a test, not a forecast. Past prices do not say what happens next.
How we ran it
The method, in full
- Prices: Bitcoin, Ether and Solana, second by second, for the whole of 12 March 2020.
- Arlos: one opened every minute of the day, at 2x, 5x, 10x, 20x, 50x and 100x, with every stop the app offers at that leverage, from 0.25% away out to the widest it allows. High leverage allows fewer stops, so it had fewer test Arlos. That makes 43,200 for each size.
- Rules: Arlo’s current rules, with the caps, the line and the cushion.
- Stops that do not fire: each stop failed to fire about 1 time in 175, the rate seen on GMX. A stop that does not fire leaves its trade open until the line closes it. The shares are averages over all the test Arlos.
- Timing: the worse of a one-second and a five-second delay for stops, and five seconds for the line.
- Sizes: the test ran at all three sizes. This page shows the size that did worst, so the shares are the same whichever size you pick. The dollar amounts follow the size you pick.