Kalshimy next prediction strategy is to buy markets at random ...just so I have a control group
my next prediction strategy is to buy markets at random ...just so I have a control group
VOXindie参与话题 14消息总数 410
该用户的聊天记录定位。
聊天记录
共 410 条，显示第 351-400 条Isn't that a losing position? lol
Isn't that a losing position? lol
$10 can work in theory, but if it hits wrong at the start it screws everything over...and as you can see there are a couple of points where the timing hits the ONLY loss in a long band of wins.
$10 can work in theory, but if it hits wrong at the start it screws everything over...and as you can see there are a couple of points where the timing hits the ONLY loss in a long band of wins.
Need to let it run for like a week to get some statistical significance.
Need to let it run for like a week to get some statistical significance.
lol I did with like $10 before, but it doesn't work unless I place bets more frequently would need like ~$120 to really test it out
lol I did with like $10 before, but it doesn't work unless I place bets more frequently would need like ~$120 to really test it out
It's way oversimplified at the moment. Doesn't even consider volume or volatility right now.
It's way oversimplified at the moment. Doesn't even consider volume or volatility right now.
lmao no idea how good the model is right now...it was doing maybe 7% gains a day in the small window I was able to get the data for before.
lmao no idea how good the model is right now...it was doing maybe 7% gains a day in the small window I was able to get the data for before.
The issue with the chart is that each dot represents a pair so it might be a little awkward to color code pairs...but should be doable.
The issue with the chart is that each dot represents a pair so it might be a little awkward to color code pairs...but should be doable.
Coding by market ticker is a good idea.
Coding by market ticker is a good idea.
I just finished moving everything over into persistent storage (SQL) so I can play more with the charts.
I just finished moving everything over into persistent storage (SQL) so I can play more with the charts.
That's what I was planning on doing next. And it's not targeting a single market. It's dynamically picking the market pairs with the highest predicted value at the moment (I assume the bands have something to do with that).
That's what I was planning on doing next. And it's not targeting a single market. It's dynamically picking the market pairs with the highest predicted value at the moment (I assume the bands have something to do with that).
The pairing isn't necessary though. In theory it would work with single bets (e.g. BTC over 62,000)
The pairing isn't necessary though. In theory it would work with single bets (e.g. BTC over 62,000)
Nope lol It's creating bet pairs that it expects to be profitable over time based on some historical data.
Nope lol It's creating bet pairs that it expects to be profitable over time based on some historical data.
They're consistently "profitable" because I'm only picking bets I predict to be profitable. If it can't find any, it'll show 0...but those tend to be less frequent.
They're consistently "profitable" because I'm only picking bets I predict to be profitable. If it can't find any, it'll show 0...but those tend to be less frequent.
Predicted is for over long periods of time. As you can see, the actual on the hourly events is a bit different lol
Predicted is for over long periods of time. As you can see, the actual on the hourly events is a bit different lol
The events are from the hourly crypto events.
The events are from the hourly crypto events.
The banding is similar in different events.
The banding is similar in different events.
Calculates fees and a 1 cent buffer so all the orders actually fill.
Calculates fees and a 1 cent buffer so all the orders actually fill.
The data should be legit. Verified that the numbers line up in a calculation from an independent/separate DB instance.
The data should be legit. Verified that the numbers line up in a calculation from an independent/separate DB instance.
Any math nerds care to explain the apparent banding I'm seeing here in my predictions/results (cyan scatterplot)? I'm not a pro with statistics.
Any math nerds care to explain the apparent banding I'm seeing here in my predictions/results (cyan scatterplot)? I'm not a pro with statistics.
as long as you don't place orders it doesn't matter
as long as you don't place orders it doesn't matter
you can test models with live prod data
you can test models with live prod data
If you're just trying to validate prediction models, you can do that without actually placing orders.
If you're just trying to validate prediction models, you can do that without actually placing orders.
Make a second account and push the demo data that you need lol
Make a second account and push the demo data that you need lol
I'm a perp 🔫🔫😎
I'm a perp 🔫🔫😎
It is in us-east-2. See my screenshot.
It is in us-east-2. See my screenshot.
They def on east-2.
They def on east-2.
I get 20ms from FL
I get 20ms from FL
Anyone know if there's a limit to how many markets you can `delete_markets` on in a single `update_subscription` WS command? I'm batching 10 at a time, but it would be better if I could go higher and it was seemingly finnicky when I upped it before.
Anyone know if there's a limit to how many markets you can `delete_markets` on in a single `update_subscription` WS command? I'm batching 10 at a time, but it would be better if I could go higher and it was seemingly finnicky when I upped it before.
You gotta get a server in the Republic of Kiribati so you're further in the future than all these clowns in Ohio.
You gotta get a server in the Republic of Kiribati so you're further in the future than all these clowns in Ohio.
Every message should be signed more or less like this: ``` private void _addAuthHeaders(HttpRequestMessage request) { // Signature = RSA-PSS( timestamp + METHOD + /path ) // Path must NOT include the base URL, just /trade-api/v2/... string timestamp = DateTimeOffset.UtcNow.ToUnixTimeMilliseconds().ToString(); string method = request.Method.Method.ToUpper(); string path = request.RequestUri!.AbsolutePath; // e.g. /trade-api/v2/portfolio/balance byte[] msgBytes = Encoding.UTF8.GetBytes($"{timestamp}{method}{path}"); byte[] sigBytes = _privateKey.SignData(msgBytes, HashAlgorithmName.SHA256, RSASignaturePadding.Pss); string signature = Convert.ToBase64String(sigBytes); request.Headers.Add("KALSHI-ACCESS-KEY", _apiKeyId); request.Headers.Add("KALSHI-ACCESS-TIMESTAMP", timestamp); request.Headers.Add("KALSHI-ACCESS-SIGNATURE", signature); } ```
Every message should be signed more or less like this: ``` private void _addAuthHeaders(HttpRequestMessage request) { // Signature = RSA-PSS( timestamp + METHOD + /path ) // Path must NOT include the base URL, just /trade-api/v2/... string timestamp = DateTimeOffset.UtcNow.ToUnixTimeMilliseconds().ToString(); string method = request.Method.Method.ToUpper(); string path = request.RequestUri!.AbsolutePath; // e.g. /trade-api/v2/portfolio/balance byte[] msgBytes = Encoding.UTF8.GetBytes($"{timestamp}{method}{path}"); byte[] sigBytes = _privateKey.SignData(msgBytes, HashAlgorithmName.SHA256, RSASignaturePadding.Pss); string signature = Convert.ToBase64String(sigBytes); request.Headers.Add("KALSHI-ACCESS-KEY", _apiKeyId); request.Headers.Add("KALSHI-ACCESS-TIMESTAMP", timestamp); request.Headers.Add("KALSHI-ACCESS-SIGNATURE", signature); } ```
Because it's only a single HTTP request. Would be more useful if I could set things like fill_or_kill to apply to the entire batch as a whole.
Because it's only a single HTTP request. Would be more useful if I could set things like fill_or_kill to apply to the entire batch as a whole.
yeah, as long as the api is in a nice format like that ...not every company does that though lol
yeah, as long as the api is in a nice format like that ...not every company does that though lol
What language is that? C++?
What language is that? C++?
I'd probably still validate it the first time you use it. Kalshi doesn't have the best error messages (at least on the web socket APIs).
I'd probably still validate it the first time you use it. Kalshi doesn't have the best error messages (at least on the web socket APIs).
If it's coming from an openapi format, it's probably generated from their actual code.
If it's coming from an openapi format, it's probably generated from their actual code.
That's on Kalshi and their QA department lol
That's on Kalshi and their QA department lol
It has a place in modern workflows, but I'm still mad sus about it...not to mention it hasn't proved it has staying power economically
It has a place in modern workflows, but I'm still mad sus about it...not to mention it hasn't proved it has staying power economically
Genned code is as good as the input lol
Genned code is as good as the input lol
Ah, sorry. Thought it was relevant to our other convo.
Ah, sorry. Thought it was relevant to our other convo.
Like I said, it can work for boilerplate, but at some point it'll miss something that'll cause you almost as long (if not longer) to debug.
Like I said, it can work for boilerplate, but at some point it'll miss something that'll cause you almost as long (if not longer) to debug.
AI's good at doing basic code review and log reading. Everything else is significantly higher quality coming from the hands of a good engineer. It can be useful for writing boilerplate, but you basically have to PR *everything* that it spits out and then correct the garbage coding styles they love to hold onto for dear life.
AI's good at doing basic code review and log reading. Everything else is significantly higher quality coming from the hands of a good engineer. It can be useful for writing boilerplate, but you basically have to PR *everything* that it spits out and then correct the garbage coding styles they love to hold onto for dear life.
lmao god no the code and unit tests I write are waaaay higher quality than anything them AI slopbots will put out
lmao god no the code and unit tests I write are waaaay higher quality than anything them AI slopbots will put out
mother fuuuu....there's a yaml file? lol
mother fuuuu....there's a yaml file? lol
I haven't dabbled in AI writing unit tests. It's bad enough with how much it screws up regular code. Can't even do data models off API docs consistently.
I haven't dabbled in AI writing unit tests. It's bad enough with how much it screws up regular code. Can't even do data models off API docs consistently.
Just like there's no covid cases if you don't test for it lol
Just like there's no covid cases if you don't test for it lol
I know companies that do payroll processing software and don't have any unit tests lol
I know companies that do payroll processing software and don't have any unit tests lol
Half of the fintech/banking world still using 15+ year old tech stacks.
Half of the fintech/banking world still using 15+ year old tech stacks.
C-suite executives are some of the dumbest people I've ever met.
C-suite executives are some of the dumbest people I've ever met.
I only use the work sub at work for their shit. They too dumb to do anything other than pay for the sub.
I only use the work sub at work for their shit. They too dumb to do anything other than pay for the sub.