456:
upon a rumor: that informer may be betting on another horse, and may be spreading rumors just so he can get better odds himself. Instead, as we have indicated, we need to evaluate our side information in the long term to see how it correlates with the outcomes of the races. This way we can determine exactly how reliable our informer is, and place our bets precisely to maximize the expected logarithm of our capital according to the Kelly criterion. Even if our informer is lying to us, we can still profit from his lies if we can find some reverse correlation between his tips and the actual race results.
1241:
an uncertain game/race/match, then compare them to the bookmaker's assessments, which usually comes in the form of odds or spreads and place the proper bet if the assessments differ sufficiently. The area of gambling where this has the most use is sports betting. Sports handicapping lends itself to information theory extremely well because of the availability of statistics. For many years noted economists have tested different mathematical theories using sports as their laboratory, with vastly differing results.
374:
1221:
the surprisal of harm from NOT getting the vaccination is only 16+2=18 bits. Whether or not you decide to get the vaccination (e.g. the monetary cost of paying for it is not included in this discussion), you can in that way at least take responsibility for a decision informed to the fact that not getting the vaccination involves more than one bit of additional risk.
96:
1252:. The underlying belief of the efficient market hypothesis is that the market will always make adjustments for any new information. Therefore no one can beat the market because they are trading on the same information from which the market adjusted. However, according to Fama, to have an efficient market three qualities need to be met:
1220:
The additive nature of this measure also comes in handy when weighing alternatives. For example, imagine that the surprisal of harm from a vaccination is 20 bits. If the surprisal of catching a disease without it is 16 bits, but the surprisal of harm from the disease if you catch it is 2 bits, then
64:
of his capital, rather than the expected profit from each bet. This is important, since in the latter case, one would be led to gamble all he had when presented with a favorable bet, and if he lost, would have no capital with which to place subsequent bets. Kelly realized that it was the logarithm
1240:
Information theory can be thought of as a way of quantifying information so as to make the best decision in the face of imperfect information. That is, how to make the best decision using only the information you have available. The point of betting is to rationally assess all relevant variables of
455:
The nature of side information is extremely finicky. We have already seen that it can affect the actual event as well as our knowledge of the outcome. Suppose we have an informer, who tells us that a certain horse is going to win. We certainly do not want to bet all our money on that horse just
1216:
The additive nature of surprisals, and one's ability to get a feel for their meaning with a handful of coins, can help one put improbable events (like winning the lottery, or having an accident) into context. For example if one out of 17 million tickets is a winner, then the surprisal of winning
1173:
This equation applies in the absence of any transaction costs or minimum bets. When these constraints apply (as they invariably do in real life), another important gambling concept comes into play: in a game with negative expected value, the gambler (or unscrupulous investor) must face a certain
1212:
For example, one might say that "the number of states equals two to the number of bits" i.e. #states = 2. Here the quantity that's measured in bits is the logarithmic information measure mentioned above. Hence there are N bits of surprisal in landing all heads on one's first toss of N coins.
81:
in a bettable event with two possible outcomes and even odds. Obviously we could double our money if we knew beforehand what the outcome of that event would be. Kelly's insight was that no matter how complicated the betting scenario is, we can use an optimum betting strategy, called the
369:{\displaystyle {\begin{aligned}I(X;Y)&=\mathbb {E} _{Y}\{D_{\mathrm {KL} }{\big (}P(X|Y)\|P(X|I){\big )}\}\\&=\mathbb {E} _{Y}\{D_{\mathrm {KL} }{\big (}P(X|{\textrm {side}}\ {\textrm {information}}\ Y)\|P(X|{\textrm {stated}}\ {\textrm {odds}}\ I){\big )}\},\end{aligned}}}
1266:
Statisticians have shown that it's the third condition which allows for information theory to be useful in sports handicapping. When everyone doesn't agree on how information will affect the outcome of the event, we get differing opinions.
451:
might be a horse that had too many oats or not enough water. The same mathematics applies in this case, because from the bookmaker's point of view, the occasional race fixing is already taken into account when he makes his odds.
610:
1072:
1209:, has applications to odds-analysis all by itself. Its two primary strengths are that surprisals: (i) reduce minuscule probabilities to numbers of manageable size, and (ii) add whenever probabilities multiply.
934:
101:
1228:
as probability = 1/2. As suggested above, this is mainly useful with small probabilities. However, Jaynes pointed out that with true-false assertions one can also define bits of evidence
1232:
as the surprisal against minus the surprisal for. This evidence in bits relates simply to the odds ratio = p/(1-p) = 2, and has advantages similar to those of self-information itself.
771:
1160:
1129:
1102:
821:
738:
707:
680:
966:
1178:
scenario. Note that even food, clothing, and shelter can be considered fixed transaction costs and thus contribute to the gambler's probability of ultimate ruin.
791:
653:
633:
27:
might be considered a formal expression of the theory of gambling. It is no surprise, therefore, that information theory has applications to games of chance.
1217:
from a single random selection is about 24 bits. Tossing 24 coins a few times might give you a feel for the surprisal of getting all heads on the first try.
1281:
980:
An important but simple relation exists between the amount of side information a gambler obtains and the expected exponential growth of his capital (Kelly):
86:, to make our money grow exponentially with whatever side information we are able to obtain. The value of this "illicit" side information is measured as
474:
1181:
This equation was the first application of
Shannon's theory of information outside its prevailing paradigm of data communications (Pierce).
986:
832:
1415:
1384:
1262:
All agree on the implications of the current information for the current price and distributions of future prices of each security
1190:
1206:
392:
1482:
1398:
Thermodynamics and
Thermostatics: An Introduction to Energy, Information and States of Matter, with Engineering Applications
1467:
1332:
1324:
78:
1477:
1249:
1248:. Random walk is a scenario where new information, prices and returns will fluctuate by chance, this is part of the
1276:
20:
65:
of the gambler's capital which is additive in sequential bets, and "to which the law of large numbers applies."
793:
th horse winning pays double the amount bet). This quantity is maximized by proportional (Kelly) gambling:
1320:
396:
54:
1202:
969:
19:
might be thought of as gambling theory applied to the world around us. The myriad applications for
1400:(D. Van Nostrand Company Inc., 24 West 40 Street, New York 18, New York, U.S.A) ASIN: B000ARSH5S.
1286:
1167:
436:
87:
42:
24:
23:
tell us precisely how to take the best guess in the face of partial information. In that sense,
16:
1472:
1380:
1175:
410:
1193:
Surprisal and evidence in bits, as logarithmic measures of probability and odds respectively.
743:
1341:
1198:
1410:
1138:
1107:
1080:
799:
716:
685:
658:
1419:
1372:
942:
83:
36:
1431:
Fama, E.F. (1970) "Efficient
Capital Markets: A Review of Theory and Independent Work",
1345:
776:
638:
618:
1461:
1452:
465:
1245:
60:
Part of Kelly's insight was to have the gambler maximize the expectation of the
1349:
1447:
1259:
All available information is costlessly available to all market participants
427:: we need to evaluate how accurate, in the long term, our side information
46:
605:{\displaystyle W(b,p)=\mathbb {E} =\sum _{i=1}^{m}p_{i}\log _{2}b_{i}o_{i}}
50:
1189:
1067:{\displaystyle \mathbb {E} \log K_{t}=\log K_{0}+\sum _{i=1}^{t}H_{i}}
929:{\displaystyle \max _{b}W(b,p)=\sum _{i}p_{i}\log _{2}o_{i}-H(p)\,}
1224:
More generally, one can relate probability p to bits of surprisal
1188:
710:
391:
is the state of the bookmaker's knowledge. This is the average
74:
41:
Kelly betting or proportional betting is an application of
1305:
1162:
is the amount of side information obtained concerning the
1412:
Sports
Betting from a Behavioral Finance Point of View
1141:
1110:
1083:
989:
945:
835:
802:
779:
746:
719:
688:
661:
641:
621:
477:
99:
1379:, 1st Edition. New York: Wiley-Interscience, 1991.
1256:
There are no transaction costs in trading securities
1244:
One theory regarding sports betting is that it is a
1174:
probability of ultimate ruin, which is known as the
1448:Statistical analysis in sports handicapping models
1154:
1123:
1096:
1066:
960:
928:
815:
785:
765:
732:
701:
682:, the proportion of wealth bet on the horse being
674:
647:
627:
604:
368:
443:might affect not just our knowledge of the event
1170:relative to the outcome of each betable event).
837:
90:relative to the outcome of the betable event:
1205:/uncertainty and whose average difference is
419:. Notice that the expectation is taken over
351:
261:
211:
161:
8:
1282:Statistical association football predictions
435:. This is a straightforward application of
356:
306:
241:
216:
186:
141:
1325:"A New Interpretation of Information Rate"
1146:
1140:
1115:
1109:
1088:
1082:
1058:
1048:
1037:
1024:
1005:
991:
990:
988:
944:
925:
904:
891:
881:
871:
840:
834:
812:
801:
778:
751:
745:
724:
718:
693:
687:
666:
660:
640:
620:
596:
586:
573:
563:
553:
542:
511:
500:
499:
476:
447:but also the event itself. For example,
431:is before we start betting real money on
387:is the outcome of the betable event, and
350:
349:
334:
333:
324:
323:
318:
291:
290:
281:
280:
275:
260:
259:
249:
248:
235:
231:
230:
210:
209:
198:
175:
160:
159:
149:
148:
135:
131:
130:
100:
98:
1307:Probability Theory: The Logic of Science
1297:
1077:for an optimal betting strategy, where
7:
1197:The logarithmic probability measure
1346:10.1002/j.1538-7305.1956.tb03809.x
439:. Note that the side information
253:
250:
153:
150:
14:
1185:Applications for self-information
415:distribution, or stated odds, on
21:logarithmic information measures
1453:DVOA as an explanatory variable
1409:Hansen, Kristen Brinch. (2006)
1310:(Cambridge U. Press, New York).
1236:Applications in games of chance
1201:or surprisal, whose average is
635:horses, the probability of the
464:Doubling rate in gambling on a
1433:Journal of Financial Economics
1377:Elements of information theory
955:
949:
922:
916:
861:
849:
532:
529:
523:
504:
493:
481:
395:, or information gain, of the
346:
319:
312:
303:
276:
269:
206:
199:
192:
183:
176:
169:
119:
107:
1:
1333:Bell System Technical Journal
401:probability distribution of
1422:(Arhus School of Business).
1250:efficient-market hypothesis
1166:th bet (in particular, the
393:Kullback–Leibler divergence
1499:
34:
1304:Jaynes, E.T. (1998/2003)
1277:Principle of indifference
1131:is the capital after the
383:is the side information,
1104:is the initial capital,
766:{\displaystyle o_{i}=2}
655:th horse winning being
1194:
1156:
1125:
1098:
1068:
1053:
962:
930:
817:
787:
767:
734:
703:
676:
649:
629:
606:
558:
370:
1483:Statistical inference
1396:Tribus, Myron (1961)
1192:
1157:
1155:{\displaystyle H_{i}}
1126:
1124:{\displaystyle K_{t}}
1099:
1097:{\displaystyle K_{0}}
1069:
1033:
963:
931:
818:
816:{\displaystyle b=p\,}
788:
768:
735:
733:{\displaystyle o_{i}}
704:
702:{\displaystyle b_{i}}
677:
675:{\displaystyle p_{i}}
650:
630:
607:
538:
371:
53:. Its discoverer was
17:Statistical inference
1468:Gambling mathematics
1139:
1108:
1081:
987:
961:{\displaystyle H(p)}
943:
833:
800:
777:
744:
717:
686:
659:
639:
619:
475:
97:
55:John Larry Kelly, Jr
1203:information entropy
970:information entropy
405:given the value of
1478:Information theory
1435:Volume 25, 383-417
1418:2018-09-20 at the
1287:Advanced NFL Stats
1195:
1168:mutual information
1152:
1121:
1094:
1064:
958:
926:
876:
845:
813:
783:
763:
730:
699:
672:
645:
625:
602:
437:Bayesian inference
366:
364:
88:mutual information
43:information theory
25:information theory
1375:, Joy A. Thomas.
867:
836:
786:{\displaystyle i}
648:{\displaystyle i}
628:{\displaystyle m}
342:
337:
332:
327:
299:
294:
289:
284:
77:is the amount of
1490:
1436:
1429:
1423:
1407:
1401:
1394:
1388:
1370:
1364:
1363:
1361:
1360:
1354:
1348:. Archived from
1329:
1317:
1311:
1302:
1199:self-information
1161:
1159:
1158:
1153:
1151:
1150:
1130:
1128:
1127:
1122:
1120:
1119:
1103:
1101:
1100:
1095:
1093:
1092:
1073:
1071:
1070:
1065:
1063:
1062:
1052:
1047:
1029:
1028:
1010:
1009:
994:
967:
965:
964:
959:
935:
933:
932:
927:
909:
908:
896:
895:
886:
885:
875:
844:
822:
820:
819:
814:
792:
790:
789:
784:
772:
770:
769:
764:
756:
755:
739:
737:
736:
731:
729:
728:
708:
706:
705:
700:
698:
697:
681:
679:
678:
673:
671:
670:
654:
652:
651:
646:
634:
632:
631:
626:
615:where there are
611:
609:
608:
603:
601:
600:
591:
590:
578:
577:
568:
567:
557:
552:
516:
515:
503:
409:relative to the
375:
373:
372:
367:
365:
355:
354:
340:
339:
338:
335:
330:
329:
328:
325:
322:
297:
296:
295:
292:
287:
286:
285:
282:
279:
265:
264:
258:
257:
256:
240:
239:
234:
222:
215:
214:
202:
179:
165:
164:
158:
157:
156:
140:
139:
134:
69:Side information
1498:
1497:
1493:
1492:
1491:
1489:
1488:
1487:
1458:
1457:
1444:
1439:
1430:
1426:
1420:Wayback Machine
1408:
1404:
1395:
1391:
1373:Thomas M. Cover
1371:
1367:
1358:
1356:
1352:
1327:
1319:
1318:
1314:
1303:
1299:
1295:
1273:
1238:
1187:
1142:
1137:
1136:
1111:
1106:
1105:
1084:
1079:
1078:
1054:
1020:
1001:
985:
984:
978:
941:
940:
900:
887:
877:
831:
830:
798:
797:
775:
774:
747:
742:
741:
720:
715:
714:
713:(payoff) being
689:
684:
683:
662:
657:
656:
637:
636:
617:
616:
592:
582:
569:
559:
507:
473:
472:
462:
363:
362:
244:
229:
220:
219:
144:
129:
122:
95:
94:
84:Kelly criterion
71:
39:
37:Kelly criterion
33:
12:
11:
5:
1496:
1494:
1486:
1485:
1480:
1475:
1470:
1460:
1459:
1456:
1455:
1450:
1443:
1442:External links
1440:
1438:
1437:
1424:
1402:
1389:
1365:
1340:(4): 917–926.
1312:
1296:
1294:
1291:
1290:
1289:
1284:
1279:
1272:
1269:
1264:
1263:
1260:
1257:
1237:
1234:
1186:
1183:
1176:gambler's ruin
1149:
1145:
1118:
1114:
1091:
1087:
1075:
1074:
1061:
1057:
1051:
1046:
1043:
1040:
1036:
1032:
1027:
1023:
1019:
1016:
1013:
1008:
1004:
1000:
997:
993:
977:
976:Expected gains
974:
957:
954:
951:
948:
937:
936:
924:
921:
918:
915:
912:
907:
903:
899:
894:
890:
884:
880:
874:
870:
866:
863:
860:
857:
854:
851:
848:
843:
839:
824:
823:
811:
808:
805:
782:
762:
759:
754:
750:
727:
723:
696:
692:
669:
665:
644:
624:
613:
612:
599:
595:
589:
585:
581:
576:
572:
566:
562:
556:
551:
548:
545:
541:
537:
534:
531:
528:
525:
522:
519:
514:
510:
506:
502:
498:
495:
492:
489:
486:
483:
480:
461:
458:
377:
376:
361:
358:
353:
348:
345:
321:
317:
314:
311:
308:
305:
302:
278:
274:
271:
268:
263:
255:
252:
247:
243:
238:
233:
228:
225:
223:
221:
218:
213:
208:
205:
201:
197:
194:
191:
188:
185:
182:
178:
174:
171:
168:
163:
155:
152:
147:
143:
138:
133:
128:
125:
123:
121:
118:
115:
112:
109:
106:
103:
102:
70:
67:
35:Main article:
32:
29:
13:
10:
9:
6:
4:
3:
2:
1495:
1484:
1481:
1479:
1476:
1474:
1471:
1469:
1466:
1465:
1463:
1454:
1451:
1449:
1446:
1445:
1441:
1434:
1428:
1425:
1421:
1417:
1414:
1413:
1406:
1403:
1399:
1393:
1390:
1386:
1385:0-471-06259-6
1382:
1378:
1374:
1369:
1366:
1355:on 2019-04-27
1351:
1347:
1343:
1339:
1335:
1334:
1326:
1322:
1316:
1313:
1309:
1308:
1301:
1298:
1292:
1288:
1285:
1283:
1280:
1278:
1275:
1274:
1270:
1268:
1261:
1258:
1255:
1254:
1253:
1251:
1247:
1242:
1235:
1233:
1231:
1227:
1222:
1218:
1214:
1210:
1208:
1207:KL-divergence
1204:
1200:
1191:
1184:
1182:
1179:
1177:
1171:
1169:
1165:
1147:
1143:
1134:
1116:
1112:
1089:
1085:
1059:
1055:
1049:
1044:
1041:
1038:
1034:
1030:
1025:
1021:
1017:
1014:
1011:
1006:
1002:
998:
995:
983:
982:
981:
975:
973:
971:
952:
946:
919:
913:
910:
905:
901:
897:
892:
888:
882:
878:
872:
868:
864:
858:
855:
852:
846:
841:
829:
828:
827:
809:
806:
803:
796:
795:
794:
780:
760:
757:
752:
748:
725:
721:
712:
694:
690:
667:
663:
642:
622:
597:
593:
587:
583:
579:
574:
570:
564:
560:
554:
549:
546:
543:
539:
535:
526:
520:
517:
512:
508:
496:
490:
487:
484:
478:
471:
470:
469:
467:
460:Doubling rate
459:
457:
453:
450:
446:
442:
438:
434:
430:
426:
422:
418:
414:
413:
408:
404:
400:
399:
394:
390:
386:
382:
359:
343:
315:
309:
300:
272:
266:
245:
236:
226:
224:
203:
195:
189:
180:
172:
166:
145:
136:
126:
124:
116:
113:
110:
104:
93:
92:
91:
89:
85:
80:
76:
68:
66:
63:
58:
56:
52:
48:
44:
38:
31:Kelly Betting
30:
28:
26:
22:
18:
1432:
1427:
1411:
1405:
1397:
1392:
1387:, Chapter 6.
1376:
1368:
1357:. Retrieved
1350:the original
1337:
1331:
1321:Kelly, J. L.
1315:
1306:
1300:
1265:
1243:
1239:
1229:
1225:
1223:
1219:
1215:
1211:
1196:
1180:
1172:
1163:
1135:th bet, and
1132:
1076:
979:
938:
825:
614:
463:
454:
448:
444:
440:
432:
428:
424:
423:rather than
420:
416:
411:
406:
402:
398:a posteriori
397:
388:
384:
380:
378:
72:
61:
59:
40:
15:
1246:random walk
293:information
1462:Categories
1359:2019-09-05
1293:References
826:for which
709:, and the
466:horse race
1035:∑
1018:
999:
911:−
898:
869:∑
580:
540:∑
518:
307:‖
187:‖
62:logarithm
47:investing
1473:Wagering
1416:Archived
1323:(1956).
1271:See also
412:a priori
51:gambling
773:if the
740:(e.g.,
79:entropy
1383:
939:where
379:where
341:
331:
326:stated
298:
288:
1353:(PDF)
1328:(PDF)
1230:ebits
1226:sbits
1381:ISBN
711:odds
468:is
336:odds
283:side
49:and
1342:doi
1015:log
996:log
968:is
889:log
838:max
571:log
509:log
75:bit
45:to
1464::
1338:35
1336:.
1330:.
972:.
73:A
57:.
1362:.
1344::
1164:i
1148:i
1144:H
1133:t
1117:t
1113:K
1090:0
1086:K
1060:i
1056:H
1050:t
1045:1
1042:=
1039:i
1031:+
1026:0
1022:K
1012:=
1007:t
1003:K
992:E
956:)
953:p
950:(
947:H
923:)
920:p
917:(
914:H
906:i
902:o
893:2
883:i
879:p
873:i
865:=
862:)
859:p
856:,
853:b
850:(
847:W
842:b
810:p
807:=
804:b
781:i
761:2
758:=
753:i
749:o
726:i
722:o
695:i
691:b
668:i
664:p
643:i
623:m
598:i
594:o
588:i
584:b
575:2
565:i
561:p
555:m
550:1
547:=
544:i
536:=
533:]
530:)
527:X
524:(
521:S
513:2
505:[
501:E
497:=
494:)
491:p
488:,
485:b
482:(
479:W
449:Y
445:X
441:Y
433:X
429:Y
425:X
421:Y
417:X
407:Y
403:X
389:I
385:X
381:Y
360:,
357:}
352:)
347:)
344:I
320:|
316:X
313:(
310:P
304:)
301:Y
277:|
273:X
270:(
267:P
262:(
254:L
251:K
246:D
242:{
237:Y
232:E
227:=
217:}
212:)
207:)
204:I
200:|
196:X
193:(
190:P
184:)
181:Y
177:|
173:X
170:(
167:P
162:(
154:L
151:K
146:D
142:{
137:Y
132:E
127:=
120:)
117:Y
114:;
111:X
108:(
105:I
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.
