forked from r8m/MarketMaker
-
Notifications
You must be signed in to change notification settings - Fork 0
/
orderbookGetMarketParam_IntensitySignal.R
289 lines (236 loc) · 9.18 KB
/
orderbookGetMarketParam_IntensitySignal.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
#' Function for calculation Market parametrs with orderbook intensity signal
library(markovchain)
library(data.table)
source('orderbookOU.R', echo=FALSE)
getMarketParams<-function(fname,
#Time frame
TFrame=10,
#Time step
deltat=0.5,
#Open position frame
MY=10,
#Open position step
deltaY=1,
#Disbalance frame
MF=10,
# Disbalance step
deltaF=0.1,
# Price min step
deltaTick=1,
#Commision
eps=0.5,
# Invenory penalization (Risk)
gamma=2,
# Max market order size in lot
dzetamax=10,
#Spread Max
SMax=5,
# Orderbook max depth
levelF=2,
deltaN=10,
NFrame=20,
byT=FALSE){
# Load Dataset
load(fname)
# Clean and Filter Data
dfdate<-format(obDT[.N,datetime], "%Y-%m-%d")
downlimit<-as.POSIXct(paste(dfdate,"10:05:00.000"))
uplimit<-as.POSIXct(paste(dfdate,"18:40:00.000"))
# Make sum of volumes on bid and ask sides with some depth
obDT[,bidCum:=rowSums(.SD),.SDcols=paste("bidvolume",1:levelF,sep="")]
obDT[,askCum:=rowSums(.SD),.SDcols=paste("askvolume",1:levelF,sep="")]
# Make intensity signal
ddt<-0.01
timedepth<-1/ddt
timeDT<-data.table(datetime=seq(downlimit, uplimit, ddt))
setkey(timeDT, datetime)
setkey(obDT, datetime)
obtDT<-obDT[timeDT,roll =T, mult="last"]
obtDT[, paste("lagdbidvol",1:timedepth,sep=""):=lapply(1:timedepth, FUN=function(i)shift(abs(bidCum-shift(bidCum, n=1, type="lag")>0),
n=i, type="lag"))]
obtDT[, intbid1:=rowSums(.SD),.SDcols=paste("lagdbidvol",1:timedepth,sep="")]
obtDT[,paste("lagdbidvol",1:timedepth,sep=""):=NULL]
obtDT[, paste("lagdaskvol",1:timedepth,sep=""):=lapply(1:timedepth, FUN=function(i)shift(abs(askCum-shift(askCum, n=1, type="lag")>0),
n=i, type="lag"))]
obtDT[, intask1:=rowSums(.SD),.SDcols=paste("lagdaskvol",1:timedepth,sep="")]
obtDT[,paste("lagdaskvol",1:timedepth,sep=""):=NULL]
obtDT[, logF:=round(log(intask1)-log(intbid1),abs(floor(log10(deltaF))))]
#ggplot(obtDT[1E5:1.1E5])+
# geom_line(aes(x=datetime, y=(askprice1+bidprice1)/2))+
# geom_point(aes(x=datetime, y=(askprice1+bidprice1)/2, color=factor(sign(logF))))
obtDT<-obtDT[, .(datetime, logF)]
obtDT<-obtDT[!is.na(logF)]
obtDT<-obtDT[!is.infinite(logF)]
obDT<-obtDT[obDT,roll =T, mult="last"]
rm(obtDT, timeDT, ddt,timeDT)
######################
#' Spread S
obDT[,deltaS:=round(askprice1-bidprice1, abs(floor(log10(deltaTick))))]
#' Volume Imbalance F
#obDT[,logF:=round(log(bidCum)-log(askCum),abs(floor(log10(deltaF))))]
#' Fair Price
obDT[,pricemid:=(askprice1+bidprice1)/2]
#' Moving average filter
#obDT[,pricemid:=EMA(pricemid,50)]
#obDT[,deltaS:=EMA(deltaS,deltaN)]
#obDT[,logF:=EMA(logF,50)]
obDT[,deltaS:=round(deltaS, abs(floor(log10(deltaTick))))]
#' Volume Imbalance F
obDT[,logF:=round(logF,abs(floor(log10(deltaF))))]
#Clean and Filter Data
SMax<-SMax*deltaTick
obDT<-obDT[deltaS<=SMax & deltaS>0,]
obDT<-obDT[datetime>downlimit & datetime<uplimit]
obDT<-obDT[complete.cases(obDT)]
tickDT<-tickDT[datetime>downlimit & datetime<uplimit]
MF<-ceiling(max(abs(obDT$logF)))
obDT<-obDT[abs(logF)<=MF]
#' Orderbook parameter Estimates:
#' TOTAL TIME
NN<-obDT[,.N]
TT<-NN/deltaN #as.numeric(difftime(obDT[.N,datetime],obDT[1,datetime], unit="secs"))
#' TOTAL NUMBER OF OBSERVATIONS
# Average time interval between bid-asks times shiftvalue
if (byT==FALSE)
{
deltat<-deltaN*TT/NN
TFrame<-round(deltat*(NFrame/deltaN),2)
}
if (byT==TRUE)
{
deltaN=round(deltat*NN/TT,0)
NFrame=deltaN*round(TFrame/deltat,0)
}
obDT[,jumpS:=shift(deltaS,deltaN,type="lead")-obDT$deltaS]
obDT<-obDT[complete.cases(obDT)]
#' Spread jump intensivity lambdaS
lambdaS<-obDT[jumpS!=0,.N]/TT
#' Spread transition matrix roS
roS<- markovchainFit(data=obDT[jumpS!=0,deltaS])
SMax=nrow(roS$estimate)*deltaTick
#' Mean reversion parameter F alfaF
#' Volatility paramter F sigmaF
ouCoef<-ou.fit (obDT$logF,deltat)
alfaF<-as.numeric(ouCoef["theta2"])
sigmaF<-as.numeric(ouCoef["theta3"])
#' Price jump intensivity lambdaJ1, lambdaJ2
obDT[,pricemidJump:=shift(pricemid, deltaN,type="lead")-pricemid]
obDT<-obDT[complete.cases(obDT)]
pmJump1<-obDT[abs(pricemidJump)>=deltaTick/2 & abs(pricemidJump)<deltaTick]
lambdaJ1<-pmJump1[,.N]/TT
pmJump2<-obDT[abs(pricemidJump)>=deltaTick]
lambdaJ2<-pmJump2[,.N]/TT
#' prob. distribution parameters of directions of mid-price jumps beta1, beta2
psi1<-data.table(table(pmJump1$logF))
names(psi1)<-c("logF", "Freq")
psi1[,logF:=as.numeric(logF)]
psi1[,Freq:=Freq/pmJump1[,.N]]
psi1[,Prob:=cumsum(Freq)]
beta1<-as.numeric(coef(glm(Prob~logF-1,data=psi1, family=quasibinomial(link = "logit")))[1])
#beta1<-1/as.numeric(coef(fitdistr(psi1$Prob, "logistic", location=0)))
psi1[,Fit:=1/(1+exp(-beta1*logF))]
psi2<-data.table(table(pmJump2$logF))
names(psi2)<-c("logF", "Freq")
psi2[,logF:=as.numeric(logF)]
psi2[,Freq:=Freq/pmJump2[,.N]]
psi2[,Prob:=cumsum(Freq)]
beta2<-as.numeric(coef(glm(Prob~logF-1,data=psi2, family=quasibinomial))[1])
#beta2<-1/as.numeric(coef(fitdistr(psi2$Prob, "logistic", location=0)))
psi2[,Fit:=1/(1+exp(-beta2*logF))]
#
# ggplot()+
# geom_point(data=psi2,aes(x=logF, y=Prob),color="mediumaquamarine")+
## geom_point(data=psi1,aes(x=logF, y=Prob),color="lightcoral")+
# geom_line(data=psi2,aes(x=logF, y=Fit),color="lightcoral")+
# ggtitle(paste("beta2 =",round(beta2,2), sep=" "))
#
# ggplot()+
# geom_point(data=psi1,aes(x=logF, y=Prob),color="mediumaquamarine")+
# geom_line(data=psi1,aes(x=logF, y=Fit),color="lightcoral")+
# ggtitle(paste("beta1 =",round(beta1,2), sep=" "))
# Market order jump intensivity at ask (lambdaMA) and bid size (lambdaMB)
setkey(tickDT, datetime)
setkey(obDT, datetime)
tbaDT<-obDT[tickDT,roll =T, mult="last"]
tbaDT<-tbaDT[complete.cases(tbaDT)]
lambdaMA<-tbaDT[,sum(price>=askprice1)]/TT
lambdaMB<-tbaDT[,sum(price<=bidprice1)]/TT
# Limit order fill rates dzeta0, dzeta1
h<-data.table(table(tbaDT[price<=tbaDT$bidprice1 | price>=tbaDT$askprice1,logF]))
names(h)<-c("logF", "Freq")
h[,logF:=as.numeric(logF)]
h[,Freq:=Freq/tbaDT[,.N]]
h[,Prob:=cumsum(Freq)]
dzeta<-as.numeric(coef(glm(Prob~logF, data=h, family=quasibinomial(link = "logit"))))
ff<-glm(Prob~logF, data=h, family=quasibinomial(link = "logit"))
#h$FitP<-predict(ff,type="response")
h$Fit<-1/(1+exp(-(dzeta[1]+dzeta[2]*h$logF)))
# ggplot()+
# geom_point(data=h,aes(x=-logF, y=Prob),color="mediumaquamarine")+
# geom_line(data=h,aes(x=-logF, y=Fit),color="lightcoral")#+
# geom_line(data=h,aes(x=logF, y=FitP),color="lightblue")
# Time Length in seconds
# Size of time step in seconds
TT<- seq(0,TFrame, by=round(deltat,2))
NT<-length(TT)
# Inventory grid bound in lot
# Inventorygrid step size in lot
YY<-seq(-MY, MY, by=deltaY)
NY<-length(YY)
# Depth imbalance grid bound
# Depth imbalance grid step size
FF<-seq(-MF, MF, by=deltaF)
NF<-length(FF)
# Tick size
# Commision
SS<-seq(deltaTick,SMax, by=deltaTick)
NS<-length(SS)
# Number of Monte Carlo simulation paths
NMC<-0
# Initial cash
X0<-0
# Initial inventory
Y0<-0
# Initial mid-price of stock
P0<-0
obMarketParam<-list(
dfdate=dfdate,
lambdaS=lambdaS,
roS=roS$estimate,
alfaF=alfaF,
sigmaF=sigmaF,
lambdaJ1=lambdaJ1,
lambdaJ2=lambdaJ2,
beta1=beta1,
beta2=beta2,
lambdaMA=lambdaMA,
lambdaMB=lambdaMB,
dzeta0=dzeta[1],
dzeta1=dzeta[2],
TFrame=TFrame,
deltat=deltat,
deltaN=deltaN,
NFrame=NFrame,
TT= TT,
NT=NT,
MY=MY,
deltaY=deltaY,
YY=YY,
NY= NY,
MF=MF,
deltaF=deltaF,
FF=FF,
NF=NF,
deltaTick=deltaTick,
eps=eps,
gamma=gamma,
dzetamax=dzetamax,
SMax=SMax,
SS=SS,
NS=NS,
NMC=NMC,
X0=X0,
Y0=Y0,
P0=0
)
}