该用户的聊天记录定位。
you can probably run several backtests with each assumption. 50/50 above/below gets removed. 100% above gets removed. 100% below gets removed. instead of wiating for whole level, i think that would be clearer
you can probably run several backtests with each assumption. 50/50 above/below gets removed. 100% above gets removed. 100% below gets removed. instead of wiating for whole level, i think that would be clearer
Now youre avoiding a lot of potential adverse selection
Now youre avoiding a lot of potential adverse selection
but either way it doesnt need to be
but either way it doesnt need to be
no longer 1:1 though
no longer 1:1 though
probably close to 50% of cancel will be in front and 50% will be behind
probably close to 50% of cancel will be in front and 50% will be behind
I think a very common case
I think a very common case
great minds think alike
great minds think alike
same exact examples
same exact examples
holy aura
holy aura
there are 100 shares at 0.4, you add 10 more. now there are 110. someone else comes(after you) and adds 100 more shares. now there are 210 shares with you in the middle. 50 shares get removed. Is your queue now 100th-110th share or 50-60th share
there are 100 shares at 0.4, you add 10 more. now there are 110. someone else comes(after you) and adds 100 more shares. now there are 210 shares with you in the middle. 50 shares get removed. Is your queue now 100th-110th share or 50-60th share
unless its not, then its not
unless its not, then its not
+EV is +EV
+EV is +EV
none the less, your testing system does not to be this rigorous
none the less, your testing system does not to be this rigorous
but thats only for some strict scenarios
but thats only for some strict scenarios
or unless if you can measure the amt added after as a multiple/inclusive of the cancel
or unless if you can measure the amt added after as a multiple/inclusive of the cancel
L3 book is the only way to model this
L3 book is the only way to model this
yes but if someone who joined below you cancels, you get more priority. if someone after, your priority is the same
yes but if someone who joined below you cancels, you get more priority. if someone after, your priority is the same
you cant just see if the amt removed is the amt added at one tick
you cant just see if the amt removed is the amt added at one tick
im assuming if they place after it will be multiple users at once
im assuming if they place after it will be multiple users at once
Say you enter a level that has 1k orders on it, then after you another 1k orders get placed. how do you determine if they were canceled before or after
Say you enter a level that has 1k orders on it, then after you another 1k orders get placed. how do you determine if they were canceled before or after
exactly
exactly
Everyday
Everyday
Trades?
Trades?
I just usually assume if it exists 250ms after i first noticed/wouldve placed then it wouldve fillewd
I just usually assume if it exists 250ms after i first noticed/wouldve placed then it wouldve fillewd
the blockchain only lists second by second not sub second fills
the blockchain only lists second by second not sub second fills
L2 im assuming
L2 im assuming
i was offering something more primative
i was offering something more primative
ahh, you record full on markets tick by tick
ahh, you record full on markets tick by tick
How did you estimate fifo queue position
How did you estimate fifo queue position
Interesting
Interesting
Much harder, but your goal would be to validate edge, not go live offline
Much harder, but your goal would be to validate edge, not go live offline
all public and easy to pull
all public and easy to pull
its probably the easiest thing to do
its probably the easiest thing to do
not hard either lmao
not hard either lmao
Yeah
Yeah
thats not hard when youre buying at 0.7
thats not hard when youre buying at 0.7
are there really so many people using ML?
are there really so many people using ML?
if your fills are +EV theyre +EV
if your fills are +EV theyre +EV
I have a pretty effective model for simulating a market making environment offline, its not hard. everything is onchain
I have a pretty effective model for simulating a market making environment offline, its not hard. everything is onchain
exactly
exactly
for making?
for making?
Probably not worth the effort isntead of creating your own but yeah
Probably not worth the effort isntead of creating your own but yeah
Youre right i may have overstated
Youre right i may have overstated
it wont tell you their model
it wont tell you their model
your reverse engineering would say: when their model disagrees they buy
your reverse engineering would say: when their model disagrees they buy
You can, but not their model
You can, but not their model
i’m all out of technical literature to read so if you’re going good i’m gonna have more
i’m all out of technical literature to read so if you’re going good i’m gonna have more
let me know how it goes
let me know how it goes
100%
100%
not alpha, but the effectivity of their alpha
not alpha, but the effectivity of their alpha
Polymarket