←
Astraeus759 posts
Astraeus avatar

Astraeus

@AstraeusCap

average crayon eater | I (try to) build and run mev bots and trade systematic crypto strategies

◷ Joined April 2021▧ 110 Photos and videos
Astraeus avatar
Astraeus@AstraeusCap·18h

Managed to decode some of the +convexity vault, a lot of nerding out in here, and then a tradeable (allegedly profitable) signal which was adapted from their rule. Had a lot of fun writing this, enjoy!

Astraeus avatar
Astraeus@AstraeusCap·Nov 17

We may be back

Post attachment

Started testing this live today (@ 4% daily vol), the first image is daily correlation of the signals Much smarter people than me have told me to heavily discount the backtest (which gave an implausibly high sharpe despite any anti-overfitting measures I took) and they might be right because it’s down 6% in the first day (approx 1/15 chance with a sharpe discounted to 1/2 of the backtest) Fun times will be had seeing how this ends up running!

Post attachmentPost attachment
Astraeus avatar
Astraeus@AstraeusCap·Nov 16

Started testing this live today (@ 4% daily vol), the first image is daily correlation of the signals Much smarter people than me have told me to heavily discount the backtest (which gave an implausibly high sharpe despite any anti-overfitting measures I took) and they might be right because it’s down 6% in the first day (approx 1/15 chance with a sharpe discounted to 1/2 of the backtest) Fun times will be had seeing how this ends up running!

Post attachmentPost attachment

Net 15 (allegedly overfit) but uncorrelated signals together and it says my sharpe is >8 we’re about to find out. test in prod

Astraeus avatar
Astraeus@AstraeusCap·Nov 16

A funny gambling product nobody has made yet is to have a live price feed and actual profitable trading strategies/signals with backtests showing the EV + win rate etc, and users can bet on one off signals (for signals computed on xyz close timing) Imagine paying $5 to tail a trade that has historically worked 70% of the time + being able to sweat the PnL of the signal live, and you get paid from the winning trade if it’s right Would be so entertaining, surprised it hasn’t been done yet

Astraeus avatar
Astraeus@AstraeusCap·Nov 15

Net 15 (allegedly overfit) but uncorrelated signals together and it says my sharpe is >8 we’re about to find out. test in prod

Astraeus avatar
Astraeus@AstraeusCap·Nov 14

List of articles coming soon (if there are any specific topics that you guys want me to do research on, let me know and i’ll add them): 1. Why stops don’t work well on mean reversion signals (it’s actually a hidden bet on momentum) 2. Why parameters aren’t edges and why robust edges don’t really care too much about the exact parameters 3. Microcap beta adjusted sensitivity to Bitcoin 4. Most crypto pumps reverse and why most of them aren’t tradeable 5. How fees can create a mirage with regards to making signals look more predictive than they are have a bunch of effects that are slowly dwindling away as well but want to keep this sort of content educational and direct because i’m sharing stuff as i’m learning it! But if you have any ideas you want me to test or any strategies you want to know how to make it better, dms are open so happy to do the research

Astraeus avatar
Astraeus@AstraeusCap·Nov 14

Had started in 23 because I think that the ETFs brought flows into the top coins and less scammy stuff started pumping as much (that’s also why I restricted my universe to top 5/10) Additionally, the way shitcoins behaved prior to 24/25 and after I’ve seen is different (in my models at least, I could be really wrong) so I wanted to model on the newer regime crypto has gotten itself into! But covid was something I should’ve modeled, didn’t have enough data. But makes sense to me because the premium has been dwindling over the years Was not able to figure out your universe rules though, and did mistakenly characterize the vault behavior during a crash as a manual intervention when it wasn’t. Lots to learn still, and more than trying to reverse engineer everything 100%, it was just cool trying to learn about different strategies that work and understanding why they work. Which I do appreciate you giving me credit and also for explaining what I missed!

Quick question! Why are you starting in 2023 and not in 2022? I would love to see how the system handle the LUNA crash and the bear market. (Survivorship bias free db) My extra layers make it survive like nothing happened. Curious to see the results!

Astraeus avatar
Astraeus@AstraeusCap·Nov 13

Helping my friends who are in my university’s investment group build a workspace for them to more easily find ideas and develop them collaboratively as a club, has been a lot of fun digging into the research to make a cool tool Has been fun adapting what I made for myself to make research faster + easier for crypto into a tradfi version. Would be nice to eventually see a trading platform that allows this to show up, think they’d take a share from the tradingview/other terminals of the world

Post attachment
Astraeus avatar
Astraeus@AstraeusCap·Nov 12

Can someone smarter than me answer please: Is trend following effectively just a synthetic long straddle where you pay some form of (what is basically) theta decay on low vol months? Been looking into other HL vaults that trend follow and this seems to be the case. Am I missing something?

Post attachment
Astraeus avatar
Astraeus@AstraeusCap·Nov 11

The vault is real, and has been live for the last 7 months: tradingstrategy.ai/vaults/fu… This is the vault for anyone interested. @FMChain_ was kind enough to reach out and let me know it was his company’s vault when he saw the PnL shape from my last post, and then discuss with me once I had more data while I was writing this up.

Astraeus avatar
Astraeus@AstraeusCap·Nov 11

As promised, here's my first attempt to reverse engineer profitable live Hyperliquid vaults (will not be the last). Put quite a bit of work into this, so I hope this is an enjoyable read. If I made mistakes, please point them out!

Astraeus avatar
Astraeus@AstraeusCap·Nov 10

Coming soon. Effectively it’s short gamma and that has paid over the years on avg

Post attachment

Spent the day reverse engineering some HL vaults with >2 sharpe. This is fun, will democratize that info soon as well

Astraeus avatar
Astraeus@AstraeusCap·Nov 9

Spent the day reverse engineering some HL vaults with >2 sharpe. This is fun, will democratize that info soon as well

Astraeus avatar
Astraeus@AstraeusCap·Nov 8

Funny enough, this was the first step in the research. Show how the trade has evolved because more people are aware of it so they get punished Ironically, that’s also the reason the edge still exists: a) very capacity constrained, b) it’s a crowd following strategy now so the more people that short the coin, the better it does (minus squeezes which punish the dumb) Funding has shifted on avg dramatically, and the prevalence of squeezes has gone up as well

Post attachment

Worth adding funding to the test. If a new listing's perp funds at -0.1% per 8h, the short pays 0.3% a day, about 9% over a 30-day hold. Spot-close backtests leave that out, and it runs most negative when everyone shorts the same listing.

Astraeus avatar
Astraeus@AstraeusCap·Nov 8

Is it free money just shorting every new exchange listing? Maybe, but you need to refine the signal quite a bit Next write up + why a stop loss doesn’t immediately solve the problem and how not to get blown out when it gets short squeezed Are the other topics / ideas that you guys have that you’d want me to do research on / write about? Happy to do so if helpful otherwise I’ll just keep sharing things I find as I go along

Post attachment