Chapkrt win ? can I : we distribution ps.iq So with iid Last time - PowerPoint PPT Presentation

Chapkrt win ? can I : we

distribution ps.iq So with iid Last time : { G) new let be . at ?g Define " hitting tines " - 1 let 2 Ci Xu . - O } , = inf { n > O : Xu - c } and To - int { n 30 - : Xn Te - - . ±um:i:÷ini÷¥÷:::: - Ta ) =/ peek PIT . - Yz p - c

- 0.49 Ex - p :f÷:f""÷÷n a - i 1,8% 100000 99900 900001000002 × 10-17+17 The $1000 but in to size increase your if keawwg odds of winning you : increase your significantly you last case , you're in the , time at $10000 a If yer bet chance ! 88% scenario → 9 , - lo a- c -

tMf coin flip game the In , we charge our can to strategy gambling " ? " do better

Dein ( Gambling Policy ) last time , let nothin from the keeping - at .EE?Y::.Imae Xu - " win " of ith of " losing " bet or - fi . ! 9,4 , - Ici - i ) where and Wi 30 Wi - - .

Play ) Exe ( Bold dollars let reach want to c . Suppose we - Xin ) min ( Xi - i , c Wi = . without going over c most you idea : bet the . can maximizes strategy this Deep theorem : IPL To < To ) .

win ) ( Double til you EI Wi . { - Ci . , - o " ' it c. - 2 = - - - and W , =L , let else O - Xs = Xz=Xs : Xy - - if - = att = - 4--1 X then , , = att - a -1+2 and -_ a - I Xz X , - , then else cut - Xy - Ys - X , = - - - - - - - a -3 - a - l 4=1 , then X Xa= a - I -2 - else - , , , = Xie - att = Xs - 31-4 ± Xy : - - - X , - a = - Xut X.at ? : Cn -4 ) - influx N , Then Xn att if check N' - you can

get this scenario we Note in , - att ) - I IP ( " Y X n - - " double til this that Question does mean c this game ? . winning strategy in gives " win a you - IE ( att ) IE ( " nm Xn ) at I - - aren't deep pockets , because Answer your no enough : .

azitzt-e.tk with integer the largest be let KEN " bdovble lil you times ) win " K play lie , can we - MEK inffnza.cn if have - X µ = att we if iuxffn > : Cal ) > k - 2K - I - 2- else Xk - - - - a . - qkfa - qk ) ( att ) - 2K ) th ⇐ ( Xk ) - I -2 - we - - - - see - 2(2q)k = att pet if than less This a is .

doesn't actually til you win double We've seen system ? will " beat " any the casino . ÷2 Wilhoit ) = at Recall Xu , = at II - i ) wi @ Ci - t ) t Wn @ Cn , - 1) Xu , t Wn ( kn = - i ) - tn ( Ci , . - , Cn where knew Wn But - we independent . . - - , Cn - i , Cn C are .

independent , 2cm - I so and are So Wn ⇐ ( Xu - it win @ Cn - t ) ) ⇐ ( Xn ) = - t ) ) ⇐ ( Wn @ en . . ) IE ( Xn t = ⇐ ( Wn ) ⇐ ( kn - t ) IE ( Xu - i ) t = - q ) IELWN )lp " IE ( Xn . . ) t - 30 - IEC # So : if a - IECX , ) then - I - - - = - p - , a > IECX , )3IECXz ) 's if petz , then - - - ( X. Is # ( Xz ) E if z , thin - as - - ' p >

if bum IE( Xn ) sa pet this gives Nate . win double til you saw , with , we contrast In " nm Xn ) - att ⇐ ( - . - tellin Xa ) ? by # Xn ) when - is MCT of EXIT , then know I ;mXn We nm Xa ) ⇐ ( " linm # ( Xn ) says - . -

( Bounded convergence Theorem ) Thou almost surely . If Ippon exists lynx . that , then IX. Is k have all for we n KEIR so " nm Xu ) E- ( nm Elk ) " = . say , for ;4 The triage inequality - X PI Dude in Xu - . - Xml ) E IE ( IX - Xn ) / I ⇐ ( X - IE ( Xn ) I . I ⇐ ( x ) = so for all NEIN have all : for e > 0 we want We - E- ( Xn ) l K E IIECX ) we get . n z N - Xal ) - a -50 s O El IX as do this shewing by . we

- Xnlwll > e } - { wer I Xlw ) Define let E > 0 An : - . . have all we .r have for we We - Xnlw ) Is et 2K Han ( w ) I Xlw ) expectations Take IEC et 2K Iancu ) ) - Xnlw ) l ) E ⇐ ( lxlw ) et 2K IPC An ) = sides both take kmsup Naw on " TYP MAN ) set 2K Pl ' 'm :P An ) - Xml ) Et 2K ⇐ ( IX " map E

- X. Cull > e ) - " must { wer : lxlw ) An limsup But - line Xnlw ) t Xlw ) with of , is the set w DNE ) { wer : hnmxnlw ) just which is . is measure zero . set hypothesis this by But Hence " TYP MANI ' 'm :P An ) ' et 2K Pl - Xml ) Et 2K ⇐ ( IX " mhm E - E E . " I'm IEHHXND fer " msn.PE/lX-Xal ) < E all s , s . S . = Off