Modification to Blum's protocol of flipping a coin over the telephone so that a player can't intentionally lose Blum's protocol, as described here, http://www.math2803.gatech.edu/wp-content/uploads/CoinFlip.pdf for flipping a coin over the telephone works well, except there is no way to ensure that Bob did not lie and say he lost at the end of the protocol. If Bob claims to have won, he must send the factorization of n to Alice, but there is no such verification for when Bob says he lost.
Is there any modification to this protocol so that Bob cannot intentionally lose? 
 A: Not really a modification of the protocol, but how's this for a coin toss based on factoring being hard:


*

*Alice chooses two large random primes $p$ and $q$. She tells Bob their product $pq$.

*Bob chooses a random bit $b\in\{0,1\}$. He sends that to Alice.

*Alice reveals $p$ and $q$.

*Bob wins if his bit is the same as bit 100 of $\min(p,q)$.

The problem with this is not how it works -- if the protocol is completed as written we end up with a random bit that neither party has a way to control.
However, what if the protocol doesn't complete? This one has the feature that Alice knows who has won before Bob does. If she doesn't like the result, she might drop the connection without sending message 3, claim that her computer crashed and she doesn't remember $p$ and $q$ anymore, or whatever.
Simply restarting the protocol in that case doesn't work -- if we assume the protocol works perfectly the second time around, Alice has just secured herself a 75% overall chance of getting her preferred result.
In order to guarantee honesty in completing the protocol, it would seem that we need a way to punish a party who stops participating in the middle of it. However, within the confines of the "coin throw" metaphor the only available way to do that would be to know what Alice's desired outcome is and dictate that she loses if she suddenly goes silent. But then, in case Bob wins, she could just as well state "I forfeit" instead of sending message 3 -- and then we're back to the situation in your link.
It seems conceivable that one could design a protocol where the two parties are about equally likely to be the first to be able to prove what the result is. But that won't really help -- if one of the players has a strong preference for one result but the other is honest/indifferent, we still get lopsided results if a player can force a restart by claiming technical difficulties.
A: First he cannot intentionally lose because in the paper the author say that  Bob has 50-50 chance to guess the property of the number that Alice chose. And Bob needs to tell Alice which property he chose (odd or even) before Alice reveal to Bob the number.
