Deep Pixels

[Deep Pixels @andydeeppixels]

@Abomination81 @TrustScore_1 Legend 💪

[Deep Pixels @andydeeppixels]

@Abomination81 :)

[Deep Pixels @andydeeppixels]

@Abomination81 @paulonoldin Are you using GTC limits seeded early for FIFO, or firing FAK at signal time? We're on Ireland EC2, ~280ms total fire cycle ~260ms is sign overhead. Seeing consistent R1 misses on BTC 5m . Paulo's showing 250ms, is there an approach that clears it?

[Deep Pixels @andydeeppixels]

@milesdeutscher Trade

[Deep Pixels @andydeeppixels]

@Notofworks7 $Titanx

[Deep Pixels @andydeeppixels]

@Investalogic ✅

[Deep Pixels @andydeeppixels]

@Abomination81 Fair enough.

[Deep Pixels @andydeeppixels]

@Abombination81 Thank you sir

[Deep Pixels @andydeeppixels]

@Abombination81 Lets Go. im ready

[Abomination @Abomination81]

This is the single most valuable post I have made to date. And that is really saying something.

I debated posting this... but fuck it. If this isn't the most bookmarked and shared post in history... well Jesus can't help you.

You backtested, you lost money, you tuned, you tweaked... lost more. Said F' this and tried something new. Only to see the same thing happen again and again.

Most failures are savable, most strategies "work" if tuned right. I'm up to 320 live working strategies, and 200 others that work, I just don't use for various reasons. I'm not a genius, but you start learning a recipe for a good bot.

------------------------------------------------------------
Some facts that make the suck, suck less

#1 Regime changes. This is an easy win. Different days do different things. M-F are retail investors, Saturday and Sunday are casuals with time off. Check any bot you have, literally the opposite thing works on weekends that works weekdays most of the time. Or just skipping weekends or weekdays all together can fix it.

#2 Time. Same idea. Certain markets behave differently... asia as a whole, sucks. So when certain countries come online, things work completely differently. Many strategies can be brute forced to work by adjusting the times it trades.

#3 Negative signals. If you have a bot and its losing money... do the opposite. That simple... oh a mean reversion that is eating shit.. flip to momentum. Oh over x bps, try under etc. If you have a test that shows oh... only a 42% wr.. oh no, how shitty.. guess what, the opposite is a 58% wr.

#4 Skip losses. Most losses cluster. Logic like (If I lose two in a row, skip the next 4 windows) can fix alot of low PNL bots. Losses and wins cluster, stopping some losses is big big.

#5 Panic Sales. I posted a csv with the % chance something will drop in price. There is a point in every window, Time vs BPS, that it will not flip. You can set a logic to dump anything FAK with 100% slippage to save some of the loss.

#6 Latency. Cant fill? Sitting on books unfilled? Heavy slippage? I promise reducing latency to the true minimum will fix it. ALL backtests assume INSTANT fills, so if youre slipping, the backtest is invalid.

#8 Same strategy, different coin and time. If it logically makes sense.. try it on everything in the 5m and 15m crypto space... BTC, SOL, ETH, XRP, BNB, DOGE, HYPE all act simmilar but different. Strategies that might not work for stable btc, might do great for psycho hype.

#9 Simple. Simple. Simple. If youre hedging, rebalancing, using more than 2 signals, splitting, merging, cross window arb, mutli window consensus... etc.. You fucked up. Ai tricks you into these complicated schemes. NONE of the top bots use that crap. Its all simple brutally well execution.

#10 Try not to win, try not to lose. You can tune anything, but its hard if youre eating shit every 5m. Work on your downside. Can you do a gtc stop loss here? Can you fire an arb there? What if we skip windows that have Z signal etc.

#11 Passive > Active. Picking the windows you enter will always work better than exiting shitty windows. Buying the right side will always be better than exiting the wrong. GTC bids will always beat out FAK for reliability. The less you have to worry about laser focused execution the better.

At this point if you're not following, you don't deserve the sweet mana from the #PolyMarket gods.

[Deep Pixels @andydeeppixels]

@Abombination81 Done ✅

[Deep Pixels @andydeeppixels]

@Abombination81 You’re a legend for this. Seriously appreciate all the work you’re putting out and sharing - it’s insanely valuable.

[Abomination @Abomination81]

There is 1 strategy that dominates the leaderboards. Every top bot in crypto is doing this strategy.

Here's exactly how it's done- warning... its long.

Market Makers, Ladder Bot, Merge Bot... various names for the same strategy.

https://t.co/RdAU99RvaV

They are easy to spot. Massive PNL, thousands of trades. Prices all over the place, in whole and fractional numbers. Usually multiple markets.

So how does it work. This is conceptually a bit difficult, so stick with me.

First, they are bidding on every single token, up and down, at 1c increments pre window. Ie 45c, 46c, 47c etc. Both directions. So the entire possible bid is covered.

When someone sells at that price, they immediately put another bid up. They are providing liquidity.

They buy hundreds and hundreds of times per window. and use a tool called merge. You can merge any up/down pair at any time on Polymarket for 1$. This immediately removes the pair and returns 1$. So if you have a 70c up and a 28c down, you get 1$ for your 98c.

But this isnt how they make money, they actually lose on the merges, so a down at 70, an up at 32... merged loses 2c. But it gets rid of shares and frees up capitol.

--------------- Now the conceptual part ---------------

Assume a price has moved to 70c, and the other price is 31c.

They are placing bids at every level below those prices... ie 69.68.67.....5,4,3,2,1 on the 70 side.

They are also placing bids at every level below 31c, ie 30,29,28,27.....3,2,1

As you can see they have MANY more bids available to buy on the 70 side than the 32. Which means that they are accumulating more winning shares.

And as they buy both, they merge.. now the merge removes a losing share for 1-2c and a winning share. But because naturally the winning side goes up, you are getting more shares. Meaning you can merge all your losing shares out, leaving you with just winning shares.

So as the price gets even higher, say 95c, they have gtc buys at 94,93,92,91... 3,2,1 etc. And gtc buys on the losing side 5c, 4c, 321.

So just with math... they are going to buy many more Winners than Losers. And each loser they do buy, gets merged out with a winner.

So at window close they are holding thousands of winners they bought anywhere from 50c to 99c, and zero losers. The loser merges cost them money, but the winners are making them 35c a share (average merged cost) and the losers lost them 2.8c a share (merged loser cost).

So youre left with a net profit of roughly 32c a share. Times thousands of shares, every 5m,15m window.

Now the hard stuff.

1. The bids are filled first in first out. So if I place my ladder before you, I get filled you dont. The window opens 24hr before the actual window start (this is all crypto windows). So the battle for Los Angeles is at THE EXACT moment the window opens. Whomever can place their 100 orders first, wins.

2. Money, this shit is expensive. Takes a minimum of about $1000 usd to test per window. And you cant just test once and change, you need to test 5-6x to see if its working.

3. Slow, each change you have to wait 24hr to test again.

4. Untestable. Since its all fills against buyers. You cannot backtest this. You learn by doing.

5. the ladder isnt 1:1, you arent placing 1 share at each level, its a quadriatic "ish" curve. Ie more shares where its more likely to be at. THIS is the magic formula that takes money to test and is very difficult to get right.

6. Latency. The fight is for that first micro second to place the bids. Me and probably 25 others are fighting for that 1/25,000th of a second. You win some, you lose some.

7. Reversals. Overall your wr is very high, because youre basically DCA'ing in the entire time, BUT every once in awhile the trend reverses last second... and youre left holding 30,000 up shares naked... and it flips down. Those days... are fun.

So thats it. The entire pickle. I HIGHLY recommend you avoid this strategy. It took me $50,000 to figure it out over a month. This should be the holy grail you attempt after you have a minimum of 100 profitable bots.

If you read all this reply with - Green. Testing if this long format breakdown is worth it or if its better to do short snippets

[Deep Pixels @andydeeppixels]

@Abombination81 Claude Code + Python on a AWS EC2. We’ve tried wallet-decoding, oracle, maker, and sweep-style stuff, but most of it failed live - economics/fillability. What actually works on 15m without hidden inventory or queue edge. Momentum, mean reversion, or market-structure based?

[Deep Pixels @andydeeppixels]

@Abombination81 Done ✅

[Abomination @Abomination81]

YOUR POLYMARKET WEBSOCKET SUCKS

HERE’S HOW TO MAKE IT GOD-TIER

All bots need websockets. Websockets are the data streams from polymarket, and external API's that give us the info we need to trade.

99.99% of people (yeah, thats you) hook them up and trade. The real issues is ws are dirty, and imperfect. They miss ticks, they give stale data, they do brief disconnects etc.

This means your strategy isnt executed as you designed. You get filled worse, your hedge sucks, your entry point is missed. A perfected WS connection vs raw from polymarket can increase your bots profitability MASSIVELY.

This applies to EVERY ws you use. CLOB, Gamma, binance, ESPN, you name it.

Here is how you make them 1000x better

Layer 1.

WS warmup. Before you trade on it, you must warm it up. The first 3-5s of a ws connection is messy full of stale bids. You should start every ws connection 15s before the window opens (on up and down tokens)

Add logic that monitors tic quality for the last 5s before the window opens. You must have >=3 tics per token up/down and no single jump in >5c.

If it does have less than 3 tics or jumps more than 5, you skip that window. Your connection sucks.

Layer 2.

VOLUME. You need to dynamically spawn multiple websockets. You monitor performance and back off if it becomes unstable, but otherwise you spawn as many as your computer can handle.. For me that is 100-300 per websocket.

You have logic that creates losers. Every 4 seconds respawn the slowest 10%

Your bot takes the first tic with dedupe. So whichever 300 gives the next tic first, you use, tic by tic.

Layer 3.

Stale tic guard. Every tick is compared against the warmup monitors last knon price.
If the delta > 15c > reject it with the log (STALE TIC REJECTED)

Layer 4.

First tic skip. Every new ws connection drops its first tick (the stale order-books snapshot from PolyMarket's (or whoever's ws it is) cache.

Layer 5.

Timing offset. Do not start all 100-300 ws at once. You have logic that divides up your stable ws equally over 1 second. So if I have 100 stable, we will start up each ws in a 10ms stagger. That gives them the best chance to get different tics.

Layer 6.

Anti-jitter Reaper. Use Jitter EMA to track per connection timing variance - erratic connections get culled first. Add an 8s grace period before new ws can be culled. Have an action budget, of a max of 20 respawns/min, max 2 per cycle. Only culled replicas lose their tracking data.

I know this is a lot of text but unless every websocket, on every bot is doing this. Other traders will rip you apart.

#polymarket