Open halcyone215 opened 3 years ago
It is the mid-price at t+5. Essentially, this step is used to take latency into account by delaying your enter time point. You can see better results using the price at t, but in practice, it is unlikely to fill the order at this price.
From: halcyone215 notifications@github.com Sent: Monday, January 4, 2021 2:54 PM To: zcakhaa/DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books@noreply.github.com Cc: Subscribed subscribed@noreply.github.com Subject: [zcakhaa/DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books] Which price do you use in the Simple Trading Simulation (#3)
'Suppose our model produces a prediction of +1 at time t, we then buy shares at time t+5(taking slippage into account), and hold until -1 appears to sell all u shares (we do nothing if 0 appears).' Would you like to tell which price you use at time t+5? the mid price at t+5 or others? and why do action buy at t+5, not at t?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/zcakhaa/DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books/issues/3, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AFN66TDNXJBR3GJTCDFPKZTSYHJCPANCNFSM4VTD5IJA.
Thanks a lot for your help. Other questions confuse me are as following. 'hold until -1 appears to sell all u shares'. When and at which price you sell all u shares? at the time predict -1 appears and use the mid price at the time? And what is your train and test step time, a ticket of quotation arrived or the prediction horizon k. Is k=20 means 20 tickets of quotation?
Selling is essentially the same as buying. Suppose you get a selling signal at t, we wait for 5 more steps to account for latency and use mid-price at t+5. Time steps are tick time where order books are updated. I am not sure what you mean by the time step in terms of k, but the prediction horizon is defined in the paper.
From: halcyone215 notifications@github.com Sent: Tuesday, January 5, 2021 3:33 PM To: zcakhaa/DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books@noreply.github.com Cc: Zihao Zhang zhangzihao@hotmail.co.uk; Comment comment@noreply.github.com Subject: Re: [zcakhaa/DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books] Which price do you use in the Simple Trading Simulation (#3)
Thanks a lot for your help. Other questions confuse me are as following. 'hold until -1 appears to sell all u shares'. When and at which price you sell all u shares? at the time predict -1 appears and use the mid price at the time? And what is your train and test step time, a ticket of quotation arrived or the prediction horizon k. Is k=20 means 20 tickets of quotation?
— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/zcakhaa/DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books/issues/3#issuecomment-754709705, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AFN66TAVN34QET2ZTQBFEDLSYMWLJANCNFSM4VTD5IJA.
Thanks for your reply. For example k=20, the model give a buying signal at t, we get u shares, at t+1, another buying signal, ...at t+6,still buying signal,then we will have 6 u shares. at t+7, the model give a selling signal, we sell all 6 u shares . At time t the model give a buying signal, that is to say, it predict the price will rise at time t+k,i.e t+20,relavive to the price at t. But we do not sell the first u shares at t+20, we sell at t+7. So I think whether the model predict time step is k, the next step is t+k,not t+1. and then when sell shares, they all past the prediction horizon.
I see what you mean. It is a bit different. Suppose you get a buying signal (+1) at t, you wait until t+5 and use that mid-price as the buying price, then you hold until the -1 appears and then sell your shares (not at t+6). If you get +1 signals either at t+6 or t+7 or ..., there is no point selling your shares when the market moves in favor of you.
From: halcyone215 notifications@github.com Sent: Wednesday, January 6, 2021 12:10 AM To: zcakhaa/DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books@noreply.github.com Cc: Zihao Zhang zhangzihao@hotmail.co.uk; Comment comment@noreply.github.com Subject: Re: [zcakhaa/DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books] Which price do you use in the Simple Trading Simulation (#3)
Thanks for your reply. For example k=20, the model give a buying signal at t, we get u shares, at t+1, another buying signal, ...at t+6,still buying signal,then we will have 6 u shares. at t+7, the model give a selling signal, we sell all 6 u shares . At time t the model give a buying signal, that is to say, it predict the price will rise at time t+k,i.e t+20,relavive to the price at t. But we do not sell the first u shares at t+20, we sell at t+7. So I think whether the model predict time step is k, the next step is t+k,not t+1. and then when sell shares, they all past the prediction horizon.
— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/zcakhaa/DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books/issues/3#issuecomment-754977641, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AFN66TGG3HL74UZHNK5PAVTSYOTAPANCNFSM4VTD5IJA.
yes. As the label is the mean price for the k prediction horizon, not at the t+k time. How do you set the threshold to make the label (-1,0,1)?
I am confused. It is a classification problem where we predict +1, 0, or -1. We are not predicting prices. You can also look at https://arxiv.org/abs/2010.13924 for the setup.
From: halcyone215 notifications@github.com Sent: Wednesday, January 6, 2021 2:28 PM To: zcakhaa/DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books@noreply.github.com Cc: Zihao Zhang zhangzihao@hotmail.co.uk; Comment comment@noreply.github.com Subject: Re: [zcakhaa/DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books] Which price do you use in the Simple Trading Simulation (#3)
yes. As the label is the mean price for the k prediction horizon, not at the t+k time. How do you set the threshold to make the label (-1,0,1)?
— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/zcakhaa/DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books/issues/3#issuecomment-755329769, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AFN66TDHH6RCZWOSCBI5YZTSYRXQBANCNFSM4VTD5IJA.
Yse,your paper is a classification predict. What I say the mean price for the k prediction horizon, is that using M+(t)the mean price to get the percentage change (lt), not to say your paper is to predict price. Thanks for recommend paper.
'Suppose our model produces a prediction of +1 at time t, we then buy shares at time t+5(taking slippage into account), and hold until -1 appears to sell all u shares (we do nothing if 0 appears).'
Would you like to tell which price you use at time t+5? the mid price at t+5 or others? and why do action buy at t+5, not at t?