Thread by @duganist

A lot of the things about the BitMex swap that came to be copied widely were based on historical accidents. The Bitfinex hack caused them to adopt the round number of 1 bps per 8 hrs. Before they had a formula. They simply chose +/- 5 as well, it was a round number bias.

Now I don't dispute that those were good product design moves in 2016. This was first in class. But now there is a whole industry of perp. arbers. We're smart and we like edges. So I wanted to give people edges structurally the way Deribit did, but from a new angle.

My idea is that the base rate is based on relative interest rates like BitMex did. This will have to be an unlocked feature for the long-term future and not the default version. Once we have margin lending tx we can benchmark how people are loaning synthetic BTCs and USDs.

For the +/- 5 bps band I was thinking of using Average True Range over the preceeding hour divided by... two? By three? By four? One of those? This requires testnet experimentation! So I'll be doing some of that in Q2.

What's interesting about this is that the ATR will go up and mute the interest rate volatility while volatility in the rest of the complex goes up and drives up rates. This lower rate on our perps would lead to more premium and close the rate gap a bit.

So you can see there's a breathing cycle where being long TL perps with variable fixing bands and short e.g. Deribit in those apex moments would be good, then you'd unwind and maybe do another spread.

It's arb city.

Downside is: predictability is nice.

All this can be modeled

I think to make it more predictable the base interest rate formula and the bands would be based on the lending VWAP diff. and the ATR of the previous 1 hour session.

It would look like this probably. The 60 min ATR on the 1 min. chart. LTC is avg. 2.5 min blocks and BTC is 10 so it'd be dividing by those to get this ATR line here. Then divide that by 2, so if it's 60 (~9.5 bps) then the session band would be +/- 4.5.

This is smarter.