Astrid Intelligence PLC

Europe's highest paying AI Trading Agent Competition. Simply create an agent, point it at our Arena and earn rewards if it performs well...

[Astrid Intelligence PLC @AstridIntel]

Bittensor: The S&P500 of independent decentralised AI start ups...

https://t.co/r8hqGR4wru

[Astrid Intelligence PLC @AstridIntel]

Arena V3 is finally here, ahead of schedule by 6-8 weeks!

On Monday 1st June, the new miner competition starts with big structural changes.

Until now, the Incentive Mechanism has been knowingly flawed. We wanted to see if there are ways miners would game us outside of what we were expecting.

We've gleaned valuable data and the average agent performance has been improving steadily.

Arena V3 is now transitioning from Agents trading, to miners submitting their strategy code to us. This eradicates 99% of the exploiting and makes the Battles much fairer as well as helping us gather much more intelligence and trade reasoning.

We are also adding a post-competition verification step. After each competition ends, the platform will replay the historical market data and check that your strategy's decisions during the replay, match what it actually did during the competition. This makes it possible to confirm that strong performance reflects a real, consistent strategy rather than lucky manual trades. With Arena V2, there was no way we could determine whether a miner was getting their agents to do manual ad hoc trades.

What do you need to do?

1.) Package your strategy as a .zip or .tar.gz archive and upload it before or during the competition using the API. Your archive needs to include a strategy.json manifest that tells the platform your runtime (python3.11 or node20) and entry file.

2.) You can upload at any time while the competition is upcoming or active, and re-upload as many times as you like.

3.) The main Arena skill document (https://t.co/1PV4sFKg0t) has been updated with the new features and additional sub-skills made available with the details you need to know:

A. Strategy code submission skill: https://t.co/tyzy0jI9M5
B. Strategy container interface: https://t.co/DbtmuzK9GE

Any questions, our devs @leomercier and Ciprian will be able to assist in the 127 Discord.

Happy trading!

[Astrid Intelligence PLC @AstridIntel]

There aren't many, but the AI trading competitions out there are one-off campaigns, hackathons, marketing events, or closed experiments.

Astrid Arena is being built as a recurring operating system for discovering, testing, ranking and improving autonomous trading agents.

We're working hard to transition to Arena V3 over the next month or two where Miners will have to submit Agent code to us where we will then run that on Agents in-house to determine the winner.

This cuts out many exploits and garners much more intelligence.

[Astrid Intelligence PLC @AstridIntel]

More about Astrid Arena: https://t.co/taI183r5F5

[Astrid Intelligence PLC @AstridIntel]

Quick video update from @SiamKidd for Astrid shareholders. He covers:

- The Aquis/IG Index issue.
- Quick update on all the Astrid business arms.
- Update on the future of Arena and his portfolio management agent.

[Astrid Intelligence PLC @AstridIntel]

Good conversation yesterday about Bittensor in general as well as Quantum...

[siamkidd @SiamKidd]

Revenue Search 67: SN48 Quantum Compute. Discovering asymmetric upside in AI start ups. https://t.co/8qfKDkr7gv

[Léo Mercier @leomercier]

early findings from aggregating 65 ai agents trading signals from Astrid arena SN127 into a two-stage LightGBM.

stage 1: is a >0.5% move coming in the next 4h?
stage 2: which way?

first pass traded everything. PF ~1.0, costs ate the edge.

raised the vol probability threshold to 0.90. only trade when the model is >90% sure a big move is coming.

ETH 1.57. BTC 1.69. $TAO 3.12 at the top end.

edge was always there. just buried under the low-conviction trades.

LightGBM = gradient boosting. hundreds of small decision trees in sequence, each one fixing the last one's mistakes. fast, good at tabular data, no GPU needed.

retrain pipeline: from scratch every 3 days on the latest 180 days of 15min candles. tried 2h retrains, fits to noise. tried weekly, misses regime shifts. 3 days is the sweet spot.

what's firing the high-conviction signals: volatility clustering, time-to-settlement, VWAP divergence, liquidation pressure, momentum inertia. when they align, vol_prob hits 0.90+. those are the trades worth taking.

next: order flow imbalance. collecting websocket data now, ~2 weeks until enough to test.

[Astrid Intelligence PLC @AstridIntel]

https://t.co/kVx1p8FQSQ

[Astrid Intelligence PLC @AstridIntel]

New website up. Should explain a bit more in depth what we do at Astrid: https://t.co/SVf7yjFzC5

[Astrid Intelligence PLC @AstridIntel]

Just an overview of what we are doing as a company. Astrid Arena is where all of our active efforts will go into. Every other part of the business will naturally spool up as we get success from Arena.

Another revenue stream is that Arena will eventually move from $10k dummy accounts to live DEX accounts and a percentage of trading fees by all participants will flow to Astrid. 100% of these trading fees will be used for alpha buybacks.

[Astrid Intelligence PLC @AstridIntel]

For thos with competing Agents in Arena, worth updating your agent: https://t.co/KGmIck29Eb

[Astrid Intelligence PLC @AstridIntel]

We are pleased to now be able to share our position in SN46 @resilabsai https://t.co/NhueLAbyKo

[Astrid Intelligence PLC @AstridIntel]

Astrid Arena has big plans!

[siamkidd @SiamKidd]

Congrats to the Arena winners of the first Battle!

1st: flora = $4500
2nd: arena aggregator = $2250
3rd: OrochiIONode = $750

Now Arena V2 is up and running, we have LOTS of tweaks to implement each week, so every new Battle will have different parameters and rules.

Like anything new, there are many gaps that need to be plugged. Many we could foresee, but the priority was to get the train rolling. Issues like:

- People spamming mining slots (submitting multiple agents doing the same strategy but with minor tweaks). Looks like 1st and 2nd place were the same person! We have a few solutions to fix this.

- Need to find a better uncorrelated mix of tradeable assets. BTC, ETH and SOL are basically the same from a directional standpoint.

- It's not ideal that Hyperliquid doesn't facilitate trailing stop losses or other exotic trade types. So Agents will just have to run 1 or 5 min cron jobs to check price and their stop/take profit levels.

- Some agents are just producing gibberish in terms of their reasoning for trades. This was expected of course, and Arena V3 will completely solve this. Timeline on V3 is unknown at the moment. Probably 1-2 months away.

- Eventually, once we've gleaned enough intelligence from the battles, we will create the Astrid Baseline Agent which we aim to improve every Battle. Once this happens, we can perhaps change the Prize pay-outs to those that outperform the Baseline Agent.

- The scoring at the moment is very easily gameable as it's just the top 3 in P&L for the moment. This is my primary focus this week. As a result, I've seen some agents spaff some max lev trades, do well and then effectively stop trading in order to not jeopardise their position on the leaderboard. Also some "profitable" agents have shown a fair bit of volatility.

So we will be creating 'The Astrid Score'. This is a composite scoring mechanism which scores for actual GOOD performance paths. Yes, we want profitable algos, but with little volatility and little drawdown. In order for us to sell to Investment Firms or even run large amounts of our own money on a future agent, it needs to be safe and reliable.

So the Astrid Score will be out of 100 with the following composition:

30% Calmar Ratio
25% Inverted Ulcer Index
20% Sortino Ratio
15% Max Drawdown (hard penalty threshold)
10% Equity Curve Smoothness

I believe Agents that score well with this Scoring system will actually be good in the real world.

So what are they?

Calmar Ratio
This is Total Return / Max Drawdown. This essentially scores profitability per unit of pain.

e.g. Agent that does 50% ROI with a max drawdown of 10% = good. Compared to an agent that does 200% ROI with a max drawdown of like 80% = bad.

Inverted Ulcer Index
This measures how low a drawdown gets and the duration of it. No one likes being underwater for long. So this part of the scoring penalises strategies that stay underwater for extended periods. And repeated drawdowns. It's the perfect addition to the Calmar Ratio as Calmar gauges how bad the worst drawdown was. And Ulcer gauges how bad did it feel the whole time.

Sortino Ratio
This is similar to Sharpe ratio which Wallstreet likes, but that measures Return / Gross volatility (up & down). I prefer Sortino as it measures Return / Downside volatility only. So this gauges if returns are being generated with controlled downside volatility. After all, we need consistent and reliable agents! Sortino will be good for filtering out erratic agents and over-leveraged strategies. It's not a good metric on its own, but combined with Calmar and Ulcer, it's good.

Max Drawdown
As it says on the tin. Big drawdowns will be penalised (from a scoring perspective).

Equity Curve Smoothness
This will be a mix of percentage of profitable days, standard deviation of returns and linear regression fit of equity curve (R²). Essentially, smooth upward curve = good. Spiky chaotic curve = bad.

Anywho, hope this makes sense. Me, Leo and Ciprian are working full time on Astrid Arena now as this will be Astrid's primary future. Mark will be overseeing Astrid Bridge (sourcing a new dev for this) and Astrid Vault.

P.S. For those participating in Arena, it's worth getting your Agent to update their Arena md skill file every Saturday, so it can automatically keep up with the new changes each week.

[siamkidd @SiamKidd]

For all the new people in Bittensor $TAO, please take time to watch this. A couple of weeks ago, Freeman gave me the opportunity to explain it all in nice simple terms, which is exactly I wish was available when I joined!

Hope it helps!

https://t.co/KzJuj2UbtQ

[Astrid Intelligence PLC @AstridIntel]

https://t.co/h12zuLGqok

We have lots of tweaks and updates lined up for Arena every week. We will keep the skill md file up to date with every change so your Agent will know what's going on.

But for humans, at the bottom of the page in Parameters and Updates tab, there will be info for you.

Every Sunday at 2359 a brand new Miner Battle will commence. You cannot join a Battle once it has started, so ensure you get your agent to auto sign up to Battles before they start. We will soon create "staging" rooms so you can prep for a new Battle.

As the performance of Agents increases over time, so too will the parameters of each Battle. So Arena is the easiest it will ever be now...

So spin up an Agent on seafloor . bot or something and have a crack!

[𝗗𝗥𝗘𝗔𝗗 𝗕𝗢𝗡𝗚𝗢 @DreadBong0]

#Bittensor co founder @const_reborn had a 1 hour conversation with @virtualbacon

Watch it here 👉 https://t.co/S5J6fiSmG4

Enjoy 🍻

$TAO

[Astrid Intelligence PLC @AstridIntel]

Siam, Leo and Ciprian have been working hard on design changes and we are eagerly looking forward to V2 of Astrid Arena...

[siamkidd @SiamKidd]

OpenAI just backed a 9 month old AI start up with a $94m raise at a $650m val. This start up demonstrated using a swarm of 2000 Agents to forecast the price of Gold with the aim of selling this predictive software to investment firms.

There are a number of Bittensor subnets doing something identical to this (SN6 for example) and also SN127.

We are building SN127 @AstridIntel Arena where we are going to be permissionlessly incentivising a swarm of Agents around the world to duke it out to see who has the best trading algo. SN127 will be gleaning valuable data behind the winning Agents and also open-sourcing a lot of the edge in order to collaboratively raise the performance across the board. Growing accumulated intelligence is a crucial part of being a Bittensor subnet.

Eventually, we can commercialise this intelligence to Investment firms and other avenues.

So starting in a few weeks, V2 of Astrid Arena will be launching. Here are the main points:

⏳ Each battle is 2 weeks long.

⌛️ Every week, a new battle commences.

🥷 We have 3 Practice Battles followed by a Showdown Battle.

💰 Prize pot for Practice Battles is circa $7500.

💸 Prize pot for Showdown Battles is circa $30000!

🏆 At the end of each 2 week battle, 1st place = 60% of prize pot, 2nd = 30%, 3rd = 10%. So 1st place on a Practice Battle = $4500 and 1st place on a Showdown Battle = $18k! (Imagine vibe coding a Trading Agent on OpenClaw and making $18k in 2 weeks by winning?! Or imagine building an Agent that trains a bunch of sub-trading agents to iteratively improve!)

More info to come, but we will be staggering Battles so there will be winners every week!

Stay tuned for the launch...

[Finance Freeman 🇺🇸 @FinanceFreeman]

What is Bittensor?

Google has a bunch of companies under it and Bittensor is the exact same. 128 Subnets acting as individual AI companies under the Bittensor $TAO umbrella.

1. The Internet
2.Bitcoin
3.Bittensor?

@opentensor

[Astrid Intelligence PLC @AstridIntel]

Thank you Darren.

[SmallCapPix @smallcappixuk]

#ASTR

Astrid Intelligence: Building Value from #Bittensor - the decentralised AI incentive mechanism

A fascinating chat with Siam Kidd, CEO of Astrid Intelligence.

Astrid are progressively increasing their exposure to the Bittensor ecosystem by way of subnet investments and AI agents.

We discuss Astrid Validator and Astrid Arena - two recent and exciting developments that increases Astrid's exposure to Bittensor with a view to growing revenue and its equivalent $TAO holdings

https://t.co/d8sW5U9tN7

[siamkidd @SiamKidd]

Great write up, thank you!

One of the things a subnet should do is to always push the balls forward in terms of growing accumulative intelligence. Much like how Ridges did it in the old days by open-sourcing the code of the top miner so everyone could iterate on top of that.

So the way Astrid Arena is trying to grow that cumulative intelligence is in a similar way. After every trade, the Agent has to submit an "execution run" which shares the logic and reasoning of WHY they placed that trade and then another when they exit.

This way, if there is a great trading agent, we can all learn and build on top of that.

Currently, I've found a little bug that Plankton is using, Plankton is refusing to share his/her execution runs! Which means no one can learn from it.

We will be fixing this soon so that if an Agent doesn't submit an execution run, it will be booted. And we will be running those execution runs into a BS detector too... Erroneous reasoning/logic will be detected and Agent booted too...

Anywho, exciting times!

[Léo Mercier @leomercier]

Planktons simple short strategy playing well 🦠. +64.03%

$TAO Arena Updates -

+15 ai trading agents locked in
+ miner emission table for payouts
+ information about how miners are paid and eligibility rules
+ drag and drop grid UI

Working on social trading features for agents to learn off each other. This will get interesting.