a Dj@s`dZddlZddlZddlZdZdZddZdd d Zd d Zd dZ ddZ e dkr\e dS)aWhat did COMPLETED books actually collect, over four months of paper trading? Today's three live books each collected less than the $1.00 they owed. Three is an anecdote. mm_paper_results.csv holds every leg the paper maker quoted since April with what it filled at and whether it settled, so the same question can be asked at n in the hundreds: - of books where all three legs filled, how many collected more than 1.00? - how much more, after fees? - and does quoting a premium over the touch actually fix it, or does it just trade completed books for unfilled ones? The last question is the one that decides whether this strategy exists. A premium always improves the books that DO complete; it only makes money if enough of them still complete. Nz'E:\soccer\research\mm_paper_results.csvgQ?cCst|d|S)N?)MAKER_FEE_RATE)pr(E:\soccer\research\mm_completed_books.py maker_feesrrc CsHz*t|}||kr|kr&nn|ndWSttfyBYdS0dSN)float TypeError ValueError)xlohivrrr_fs "rcCs|r|dddSdS)N-r)rsplit)trrrevent_of#srcCstt}ttddd}t|D]}t|d}|dpBd }|r$|sRq$t |d}|durjq$t |dpxd  d v}|rt |d nd}|||dur|n||d pd d|||<q$Wdn1s0Y|S)Nzutf-8replace)encodingerrorsZ leg_tickersideZour_askfilled0)1TruetrueZfill_pxregime?)askrfillr!) collections defaultdictdictopenRESULTScsvZ DictReaderrgetlowerrstrstrip)booksZfhrevrr#rr$rrrload's"  .r2c st}dd|D}dd|D}tt|ddt|ddt|dddt|td t|d d |s|dStd g}|D]b\}}td d|D}tdd|D}||||||dt|dddqt dd|D}t|}tdt |dd||ddd|d|ddtdd|D} td| d|dd| |dd d!d"|D} td#d| D} td$| d|dd| |dd td%t | d&d'd(td)t | d&d'd(td*|d+t| d&d,d(d-d"|D} t d.d| Drbtd/t | d&d0d(ntd1td2td3td4td5td6d7d8d9d:d;d9dd=d?D]} | dgd@}}|D]\}}tdAd|DsqԇfdBd"|D}t dCd|Dr&qt|}tdDd|D}|||dq|s`qt|}t| dEdFt|d;d9t |d&dGdF|d&dHd(qtdItdJtdKtdLdS)MNcSs"i|]\}}t|dkr||qS))len.0ebrrr =zmain..cSs,i|]$\}}tdd|Dr||qS)css|]}|dVqdSrNrr6lrrr >r:z"main...)allvaluesr5rrrr9>r:,z books quoted, z with all three legs, z completed (all three filled) = gY@rz.1fz% z&=== what completed books collected ===css|]}|dVqdSr$Nrr<rrrr>Ir:zmain..css|]}t|dVqdSrBrr<rrrr>Jr:rrr!)r1collfeesnetr!css|]}|dVqdS)rDNrr6r0rrrr>Nr:z collected: median z.4fz p25 z p75 r3css|]}|dkrdVqdS)rrNr)r6crrrr>Rr:z above 1.00 before fees : /z (z.0fz%)cSsg|] }|dqS)rFrrGrrr Tr:zmain..css|]}|dkrdVqdS)rrNr)r6r rrrr>Ur:z PROFITABLE after fees : z mean net per book : dz+.2frIz median net per book : z TOTAL over z books : z+.1fcSs*g|]"}|D]}|d|dqqS)r$r#)r@)r6r8r=rrrrK\r:css|]}t|dkVqdS)g& .>N)abs)r6srrrr>]r:z fill vs quote: median z+.3fz8 fills land exactly at our quote (no per-leg slippage)z- === would quoting a premium have helped? ===zD Premium is added to EVERY leg at quote time. A book completes onlyzB if each leg still fills -- proxied here by the leg having filledz= with enough headroom between our ask and 1.00 to absorb it. z premium/legz>12 z books donez>11zmean netz>10total)rg?g?g?rg?g@rcss|]}|dVqdSr;rr<rrrr>mr:csg|]}|dqS)r$rr<ZpremrrrKor:css|]}|dkVqdS)gGz?Nrr6rrrrr>pr:css|]}t|VqdSr rCrSrrrr>sr:z>10.2fzc z>9.2fz>9.1fz@ NOTE: this holds completion FIXED, so it is an upper bound --zD it shows what a premium is worth IF fills survive it. Whether theyz> do is a queue question the paper log cannot answer; the livez' fill data is the only thing that can.)r2itemsprintr4maxsumr@appendlistsorted statisticsZmedianZmeananyr?)r/ZthreeZdoneZrowsr7r8rDrEnZaboverFwinZslipZprem_cZkeptZtotZlegsrIfrrRrmain;s     && ( (r`__main__)rr) __doc__r*r%r[r)rrrrr2r`__name__rrrrs E