+ yjRXRt^RIt^RIt^RIHtHt^RIHt]!]4Pt ]PP^] !] 44^RIHt^RIHt]!4'g.]!R]P$!4 R24]P&!^4] R, t] R , t] R , t] R , tR tR t^t^ t^t^tRtRtRRlt Rt!Rt"Rt#Rt$Rt%Rt&Rt'Rt(Rt)Rt*Rt+RRlt,Rt-].R8Xd ]-!4R#R#) uj Builds retrospective.json: a daily hits/misses/missed-opportunities summary for the dashboard's Retrospective panel. Hits/misses come from realized P/L on positions.json's closed_positions — real fills only, never estimated. Missed opportunities are new_buys candidates from daily_snapshots/ that were never entered, checked against a live TMX quote to see if they ran without us. Run manually: python3 generate_retrospective.py Cron (once daily, after TSX close, MDT): 5 14 * * 1-5 /usr/bin/python3 /media/raid/rshare/SwingTrader/generate_retrospective.py >> /media/raid/rshare/SwingTrader/daily_run.log 2>&1 N)date timedelta)Path)is_trading_day) get_quotez[generate_retrospective] z% is not a TSX trading day -- skippingzpositions.jsondaily_snapshotszscreen_criteria.jsonzretrospective.jsonz 2026-08-01g@z)funnel_breadth.recurring_megacap_handlingz 2026-08-10cVP4'gV#\V4;_uu_4p\P!V4uuRRR4# +'giR#;iN)existsopenjsonload)pathdefaultfs&& 8/media/raid/rshare/SwingTrader/generate_retrospective.py load_jsonr<s4 ;;== dqyy| s A  A cVPR4pVPR4pVPR4pVe VeVfR#R\W, V,^4R\W, V, ^d,^4/#)exit_price_cadentry_price_cadsharesN pl_dollarspl_pct)getround)pos exit_price entry_pricers& r realized_plrCsw''*+J''+,K''(#F[0FNeZ5?CeZ5DsJAN c...r2pVPR.4EFpVPR4'dKRVPR4RVPRR4RVPR4RVPR4RVPR4RVPR4R VPR 4R VPR R4R VPR R4/ p\V4pVfVPV4KVPV4VR ,^8dTMTPV4EK VP R RR7VP RR7WV3#)closed_positions_exampletickercompany entry_date close_daterrr setup_type close_reasonrcVR,#rxs&r#build_hits_misses..esAlOrTkeyreversecVR,#r+r,r-s&rr/r0fsa or)r2)rrappendupdatesort) positionshitsmisses incompletepentrypls& rbuild_hits_missesr?Os6!2r*D ]]-r 2 55    quuX quuY3 quu\2 quu\2 quu%67 quu%56 quuX quu\26 quu^R8  ^ :   e $  RL!A%699%@'3( II+TI: KK-K.  ##rc\P!V4pT\\R7,pTP R.4TP R.4,pTFpTP R4;'gRP 4TP 48wdKATP R4pT'gK\\P!T4pY8u;8:d T8:gKR#K R# \dR#i;i \dKi;i)F)daysopen_positionsr!r#r%r&T)r fromisoformat ValueErrorrENTRY_WINDOW_DAYSrupper) r# scan_date_strr8scan_dt window_end all_positionsr<edentry_dts &&& ralready_acted_onrMjs$$]39*;<.a mrTr1) SNAPSHOTS_DIRr setrtodaysortedglobrrCANDIDATE_TRACKING_STARTrFaddrMrCrDrArr5rMIN_DAYS_BEFORE_JUDGINGMISSED_OPP_THRESHOLD_PCTr7)r8r\ passed_okpendingseenrd snap_pathsnaprPcr# ref_pricer2recordrH days_sincequote current_pricerZs& rbuild_missed_opportunitiesrvsw!#RwF    ! !')) 5D JJLEM../CDE B'--2HH[) I(@@ ((9b)//R44ZDJJJAhFk*IY.<<>9-C{ HHSM9==6155B#79#Y155r#:155#7 F ,,Y7 ///Jf%E} ' 2 :#6x v&!'NMmi79DsJANH*7F& '*2F: 33#.x v&55#+x  f%#5x   (_K Fn KK+TK: g %%5  sJ J/.J/c...r2p\P4'gWV3#\4p\P!4p\ \P R44EFp\V/4;'g/pVPR4pV'd V\8dK@VPR.4;'g.EFp V PR4p V PR4;'gRp V PR4p V 'd\V 9gV fKZV P4V3p W9dKtVPV 4\WV4'dKRV RVRV /p\P!V4pY_, P p\#T 4pTeTPR 4fR TR &TP%T4EKTR ,p\'TT , T , ^d,^4pTTR &TTR &T\(8dRTR &TP%T4EK]T\*8dRTR &TP%T4EKRTR &TP%T4EK EK VP-RRR7WV3# \dEKi;i)aChecks excluded entries tagged recurring_megacap_handling for a move. Those entries are hand-waved past without full scoring to save research budget (see screen_criteria.json funnel_breadth.recurring_megacap_handling) -- this is the same shape as build_missed_opportunities, but starting from a deferral instead of a scored-and-passed new_buys candidate, to catch the case where the deferral itself cost something. rOrPexcludedr#reasonr%rSprice_at_deferral_cadrVrWrXrYrZr[r\r]cVR,#r_r,r-s&rr//build_deferred_megacap_review..rarTr1)rbr rcrrdrerfrrDEFERRAL_TRACKING_STARTRECURRING_MEGACAP_TAGrFrhrMrCrDrArr5rrirjr7)r8r\rkrlrmrdrnrorPrpr#ryrqr2rrrHrsrtrurZs& rbuild_deferred_megacap_reviewrsR"$RwF    ! !')) 5D JJLEM../CDE B'--2HH[) I(?? *b)//R/AhFh--2Fk*I2&@IDU<<>9-C{ HHSM9=='F  ,,Y7 ///Jf%E} ' 2 :#6x v&!'NMmi79DsJANH*7F& '*2F: 33#.x v&55#+x  f%#5x   ([0 Fj KK+TK: g %%5  sI66 JJc*\4pVPR.4VPR.4,F[pVPR4'gKVPR4'dK5VPVR,P44K] V#)rBr!r#r")rcrrhrF)r8tickersr<s& rdistinct_traded_tickersr speG ]]+R 09==ASUW3X X X 55??155#4#4 KK( ))+ ,Y Nrc.p\P!4pVPR.4EF pVPR4'gVPR4'gK4VPR4'd\P!VR,4MRpV'dW5, P MRpRUu.uFqtPV4eKVNK ppV'gKVe Wa8gKTP RRRRRVR,R TR VR, R R PV4 R VeTMR R2/4EK V# \dRpLi;iuupi)rBr"r#r&Ncodeunenriched_positionseverityhigh days_heldsummaryz is missing , z after zan unknown number ofz= days held -- exit/risk signals can't be computed accurately.) stop_loss_cad target_1_cad)rrdrrCrDrAr5join) r8 grace_dayslessonsrdr<rLrrmissings && r!build_unenriched_position_lessonsrs<G JJLE ]]+R 0 55  AEE(OO  >?eeL>Q>Qt))!L/:W[H08U%++T ?T?55811?T 7 )Y-D NN2VQx[Y{m< '0B/CD*3*?YE[\]FF  1* N! H Us$D:6D:5E E : E  E c|.pV'gV#VPR.4Uu/uFMpVPR4'gKVPR4'dK5VR,P4VbKO ppVPR/4;'g/PR.4;'g.FpVPR4;'gRP4pVPV4pVfKCVPR4'd\P!VR,4MRpV'd'\P !4V, P MRp TPRR R R RTR T R VPR 4RT RVPR 4 RV eT MR R2/4K V#uupi \dRpLi;i)rBr#r"rQexitsr%Nr&roverdue_exit_not_executedrcriticalrrPrz+ was flagged EXIT in the most recent scan (z) and is still open (?z6 days held). The recommendation has not been acted on.)rrFrrCrDrdrAr5) r8latest_snapshotrr<open_by_tickerer#rrLrs && rbuild_overdue_exit_lessonsr-sG (1 6F(K(K1 55? #$55#4 ( Q(K!!)R066B ; ;GR H N NB N%%/''R..0  ( ;  @C @U@Ut))#l*;<[_H7?TZZ\H,22D  5 * & ) /--k: (EoFYFYZeFfEgh&3<3Hic%RS<<  O, N5 H s(F&F&F&0F+F++ F;:F;c/pVPR.4EFpVPR4'dK\V4pVPR4;'g/P4FwrEVRJdK VPVR^R^R^R./4pVR;;,^, uu&VR,P VPR44VfKiVR ,^8dVR;;,^, uu&KVR ,^8gKVR;;,^, uu&K EK .pVP4FQwrHVP R R V 2R VRVR,RVR,RVR,RVR,R VR,\ 8/4KS V#)r!r" review_flagsTcountlossesgainsrr#rrpattern_flag actionable)rritems setdefaultr5MIN_SAMPLE_FOR_PATTERN_RULE) r8tallyr<r>rvalbucketpatternsstatss & rbuild_review_flag_patternsrNsi E ]]-r 2 55    ^%%/552<<>ID$%%dWa1gqR[]_,`aF 7Oq O 9  $ $QUU8_ 5~l#a'7Oq(O %)8$)$? 3H{{}  HTF+ $ %. %/ %. % * %.,GG  % Orc`VPRR4p/pVPR.4F}pVPR4;'gRP4pV'dVPR4'dKMVPV.4PVPR44K .pVP 4FpwrW\ V4^8dKVR8XdR MR pTPR R R TRTRTRTR\ V4 RV RRP V4 R2VR8XdRMR,/4Kr V#)aYCode-level check, not a prompt note the LLM is trusted to remember. Always flags 2+ open positions sharing a sector, with severity keyed off active_mode -- aggressive mode explicitly allows concentration in the day's leading sector (screen_criteria.json aggressive_overrides.trade_management .sector_concentration), but this script has no way to know which sector was "the leading one" on entry day, so it surfaces every concentration as real, counted evidence and leaves the "is this the intentional play" call to the scan (which has today's macro context) rather than silently assuming either way. active_mode conservativerBsectorr%r"r# aggressiveinforrsector_concentrationrrrz$ open positions are concentrated in z (rz). zAggressive mode allows this if it is the day's intentional leading-sector play, but still counts toward max_simultaneous_positions -- confirm that's actually why before adding another.zGCorrelated risk: a single sector move could hit multiple stops at once.)rstriprr5rlenr) r8criteriar by_sectorr<rrrrs && r"build_sector_concentration_lessonsrms7,,}n=KI ]]+R 0%%/''R..0z** VR(//h@ 1 G$??, w $rYz+). Force a full Step 3/4 reconfirmation on z" today, regardless of day-of-week.unenriched_positions overdue_exitsreview_flag_patternsrdeferred_megacap_moveddistinct_tickers_tradeddistinct_ticker_countnarrow_funnel_flagnarrow_funnel_threshold)rr CRITERIA_FILErrrrrrrNARROW_FUNNEL_TICKER_THRESHOLDr5rre)r8deferred_missedrn latest_snaprr unenrichedrr concentrationtraded narrow_funnel directiveslr<drs&& r build_lessonsrs$&I09)It,tK+11rH0"599:Z\]^J%++O5iLJ.yFM.y9H6yKM+I6FK"@@MJ  iLV V   iLS S  Z=F "   ) ^^  CK=! &.)*-8V7WXd d  \??   AgJ<'DQvYKPhK= 1W:,6I$))TUV_T`JaIbcAA {m<<=kN;KK`}T"$q)@'A&B%J]H^G_`778{mCe g  :=8= ?!6&>3v;=!#A G J rc(\\/4;'g/p\V4wrp\V4wrEp\ V4wrxp \ W4wr\ \RV44\RV44,^4p \V4\V4,p V 'd$\ \V4V , ^d,^4MRpR\P!4P4RRV RVR\V4R\V4R \V4R \V4R \V4/R VR VRVRVRVRVRRVRVRV /RV RV RR\ R2R\R R\ R2RR./ p\\ R 4;_uu_4p\"P$!VV^R!7RRR4\'R"\V4 R#\V4 R$\V4 R%\V4 R&\V4 R'V R(\V 4 R)V R*, 24R# +'giLr;i)+c32"TF qR,xK R#5irNr,).0hs& r main..s *TlOOTc32"TF qR,xK R#5irr,)rms& rrrs0Q&Q<&rN generated_atrrealized_pl_cad win_rate_pct hit_count miss_countincomplete_countmissed_opportunity_countdeferred_megacap_missed_countr9r:r;missed_opportunitiesr]pending_reviewdeferred_megacap_reviewr\rlrlessons_for_next_scannotesz9Missed-opportunity tracking only covers daily scans from z1 onward -- earlier scan history was not retained.z>A passed-on candidate is flagged "missed" once it is at least z.0fz(% above its scan-day price and at least z calendar days have passed.aOnly new_buys candidates are tracked for missed opportunities -- watchlist/excluded tickers and anything the screeners never surfaced are out of scope, EXCEPT excluded entries tagged funnel_breadth.recurring_megacap_handling (see deferred_megacap_review) -- those carry a price_cad captured at deferral time specifically so a hand-waved name that moved doesn't go unnoticed. Coverage starts 2026-08-10, when price_cad capture was added.zlessons_for_next_scan is read and applied by daily_scan_prompt.md Step 1a -- this is not just a dashboard display, it changes what the next scan does.w)indentz[generate_retrospective] hits=z misses=z incomplete=z missed_opps=z deferred_megacap_missed=z realized_pl=$z lessons=z distinct_tickers=r)rPOSITIONS_FILEr?rvrrrsumrrrd isoformatrgrjrir OUTPUT_FILEr dumpprint)r8r9r:r;r\rkrlr deferred_okdeferred_pendingrrrealized_totaln_closedwin_rateoutputrs rmainrsR."-33I0;D*!;I!FFw5RS\5]2O"2%29%N"G *T **S0Q&0Q-QQSTN4y3v;&H7?uSY)C/3TH  ..0   D F J &F +S-A  !    0$ !6GH`GabS TLMefiLjkSTkSllG H D e  1F@ k3  1 &!A&  (T 8CK=I*o&mCK=A##&#7"89&'y5J1K0LM#$;<=  ? s =H H __main__r )/__doc__r sysdatetimerrpathlibr__file__parentBASErinsertstrr tmx_quotesrrrdexitrrbrrrgrjrirErrr~r}rrr?rMrvrrrrrrrrr__name__r,rrrs8  $ H~3t9)  %djjl^3X YZHHQK(()) .. ,, ( "$ D& $6,?&DE&P6B>)X(?D6r zFr