PolymarketCan anyone else confirm that's the case? This makes me feel kind of insane because that's a pretty massive aggregation.
Can anyone else confirm that's the case? This makes me feel kind of insane because that's a pretty massive aggregation.
JMF参与话题 1消息总数 38
该用户的聊天记录定位。
聊天记录
共 38 条，显示第 1-38 条Ahh I think I see what's going on. The telonex is showing 23.5 shares at $0.39 but on chain it was really like 5 partial fills in that range with a range of prices. Wow that really fucking sucks for backtesting. My live $ overfills compared to using the data for backtesting because it doesn't see that a $0.40 trade was part of that 23.5 shares.
Ahh I think I see what's going on. The telonex is showing 23.5 shares at $0.39 but on chain it was really like 5 partial fills in that range with a range of prices. Wow that really fucking sucks for backtesting. My live $ overfills compared to using the data for backtesting because it doesn't see that a $0.40 trade was part of that 23.5 shares.
So what I'm seeing is that my ws collector seems to match telonex trade collector. Both diverge from telonex on chain fills.
So what I'm seeing is that my ws collector seems to match telonex trade collector. Both diverge from telonex on chain fills.
The CLOB ws.
The CLOB ws.
Looks like I"m never getting an answer from <#766038567310065704> on the WS just missing trades that show up on chain. As far as I can tell it's like 10% of trades that have no WS trade printed. Does that sound right ot people? Makes backtesting really f'in difficult if we can't rely on the WS to represent what an algo would see live.
Looks like I"m never getting an answer from <#766038567310065704> on the WS just missing trades that show up on chain. As far as I can tell it's like 10% of trades that have no WS trade printed. Does that sound right ot people? Makes backtesting really f'in difficult if we can't rely on the WS to represent what an algo would see live.
Hi can anyone speak to the data loss in the WS? When I compare the websocket data to the on chain fills, it looks like as much as 10% - 15% of fills (specific to crypto 15 m markets I'm tracking) are missing. I checked my ws feed against telonex and those match. It's just the ws based trade data in both compared to the on chain fill data from telonex that's a problem. 10% - 15% seems insane and makes backtesting on ws based data almost impossible.
Hi can anyone speak to the data loss in the WS? When I compare the websocket data to the on chain fills, it looks like as much as 10% - 15% of fills (specific to crypto 15 m markets I'm tracking) are missing. I checked my ws feed against telonex and those match. It's just the ws based trade data in both compared to the on chain fill data from telonex that's a problem. 10% - 15% seems insane and makes backtesting on ws based data almost impossible.
I pair it with internal collectors but both produce the same results. The fills match 100% on the rounds that I’ve monitored. Both miss fills picked up on chain (which telonex also has).
I pair it with internal collectors but both produce the same results. The fills match 100% on the rounds that I’ve monitored. Both miss fills picked up on chain (which telonex also has).
Is it common that the ws would miss a considerable amount of trades that appear on chain. I've been comparing telonex data vs my own collector vs the telonex on chain fill data. The former two match, so I believe we're collecting everything given by the ws, but the on chain data is picking up ~10% of fills nowhere to be found in the ws.
Is it common that the ws would miss a considerable amount of trades that appear on chain. I've been comparing telonex data vs my own collector vs the telonex on chain fill data. The former two match, so I believe we're collecting everything given by the ws, but the on chain data is picking up ~10% of fills nowhere to be found in the ws.
It's the $0.74 fill and the other fills between $0.78 and $0.73 that are on chain but not on ws.
It's the $0.74 fill and the other fills between $0.78 and $0.73 that are on chain but not on ws.
If you have any interest in seeing one round around a particular fill.
If you have any interest in seeing one round around a particular fill.
I mean, I see that there are complements, but not in that rapid sweep range.
I mean, I see that there are complements, but not in that rapid sweep range.
And I can see it in the data itself. No complements reported either.
And I can see it in the data itself. No complements reported either.
My feed matches the telonex feed, so I'm pretty confident we have the WS data that's available. My private feed obviously picks up the fill, but not the public feed.
My feed matches the telonex feed, so I'm pretty confident we have the WS data that's available. My private feed obviously picks up the fill, but not the public feed.
Yea that's what I'm looking at.
Yea that's what I'm looking at.
Looking like ~7% - 15% of fills not in the WS but on chain depending on the round.
Looking like ~7% - 15% of fills not in the WS but on chain depending on the round.
Thats what' I'm doing. Got claude handling all the analysis. The reason I caught it was that I had a real fill on my account that the telonex and my paper trader collector didn't catch, so I know it actually happened. And in that round you can see the deep sweep that got missed. It effectively dropped from like $0.8 to $0.73 in one big sweep and then recentered to $0.81 or so.
Thats what' I'm doing. Got claude handling all the analysis. The reason I caught it was that I had a real fill on my account that the telonex and my paper trader collector didn't catch, so I know it actually happened. And in that round you can see the deep sweep that got missed. It effectively dropped from like $0.8 to $0.73 in one big sweep and then recentered to $0.81 or so.
Seems to happen on deep fills that sweep a large number of prices in a short time.
Seems to happen on deep fills that sweep a large number of prices in a short time.
I'm comparing telonex to telonex in this case, but telonex seems to align with my live ws feed in terms of trade count when I include complement fills.
I'm comparing telonex to telonex in this case, but telonex seems to align with my live ws feed in terms of trade count when I include complement fills.
I'll run the analysis over a bunch of rounds to see if that holds up.
I'll run the analysis over a bunch of rounds to see if that holds up.
Yo somethings driving me a bit nuts. I'm using a WS collector and Telonex. I keep finding fills in the "on chain fills" telonex that I can't pick up on the live PM ws or on the "trades" database from telonex for a round. On my last analysis it was 20% of trades that weren't recorded. Does that make any sense? Can the WS trade feed really be that bad?
Yo somethings driving me a bit nuts. I'm using a WS collector and Telonex. I keep finding fills in the "on chain fills" telonex that I can't pick up on the live PM ws or on the "trades" database from telonex for a round. On my last analysis it was 20% of trades that weren't recorded. Does that make any sense? Can the WS trade feed really be that bad?
Or do you guys use a higher frequency chainlink stream? The basic one seems quite slow.
Or do you guys use a higher frequency chainlink stream? The basic one seems quite slow.
How do you guys deal with chainlink having significantly different price points than binance or coinbase when it comes to datafeeds? I'm running open on chainlink and spot on the rapid binance WS, but it looks like binance is consistenly running ~0.14% over chainlink and it completely botches the option pricing models.
How do you guys deal with chainlink having significantly different price points than binance or coinbase when it comes to datafeeds? I'm running open on chainlink and spot on the rapid binance WS, but it looks like binance is consistenly running ~0.14% over chainlink and it completely botches the option pricing models.
Like something that exactly matches polymarkets open price.
Like something that exactly matches polymarkets open price.
What is the best source of open price that's live and usable immediately in a round?
What is the best source of open price that's live and usable immediately in a round?
Ok so I see why you say just translate everything to up or down.
Ok so I see why you say just translate everything to up or down.
Ok clarify something for me: If there is a resting limit order to sell at $0.52 and someone buys at $0.52, will it always fill that resting order first? If there is no resting order will it literally fill a $0.48 resting limit order instead?
Ok clarify something for me: If there is a resting limit order to sell at $0.52 and someone buys at $0.52, will it always fill that resting order first? If there is no resting order will it literally fill a $0.48 resting limit order instead?
Fuck me. When I did full complements in the past it created a bunch of trades that didn't exist in reality.
Fuck me. When I did full complements in the past it created a bunch of trades that didn't exist in reality.
I meant I was replying to techno not you
I meant I was replying to techno not you
No sorry Techno not you szhizhi
No sorry Techno not you szhizhi
What do you mean by that?
What do you mean by that?
Because I don't want my paper trader to "assume" the complement because that creates a lot of extra fills in bac ktests.
Because I don't want my paper trader to "assume" the complement because that creates a lot of extra fills in bac ktests.
And how can we decipher that in the WS data?
And how can we decipher that in the WS data?
I get this at a high level, I'm trying to reconcile data coming from the Websocket. I can't seem to "find" the trade that filled my $0.58 live $ limit order. I just saw a $0.42 buy on the opposing token come through the websocket.
I get this at a high level, I'm trying to reconcile data coming from the Websocket. I can't seem to "find" the trade that filled my $0.58 live $ limit order. I just saw a $0.42 buy on the opposing token come through the websocket.
When I complement every trade in the data it creates trades that truely don't exist.
When I complement every trade in the data it creates trades that truely don't exist.
Wait so every single $0.52 UP trade also represents a $0.48 Down trade of the same number of units? that makes no sense because some times the $0.52 Up buy just eats a $0.52 Up limit order.
Wait so every single $0.52 UP trade also represents a $0.48 Down trade of the same number of units? that makes no sense because some times the $0.52 Up buy just eats a $0.52 Up limit order.
The up and down tokens trade independently on the crypto 5/15 mins. That's what I'm trying to figure out.
The up and down tokens trade independently on the crypto 5/15 mins. That's what I'm trying to figure out.
But not always the same, right?
But not always the same, right?
Can someone help me understand the mirror thing on trades? When we subscribe to the polymarket WS it seems that some trades are Up and Down independent and some are "mirrored"? So here's what I saw: 1) My live $ trader filled at $0.58 Up 2) My paper trader never saw a fill on the PM WS at $0.58 up but did see $0.42 Down
Can someone help me understand the mirror thing on trades? When we subscribe to the polymarket WS it seems that some trades are Up and Down independent and some are "mirrored"? So here's what I saw: 1) My live $ trader filled at $0.58 Up 2) My paper trader never saw a fill on the PM WS at $0.58 up but did see $0.42 Down