Will_Newsome pointed out the caveat that it's only good to admit errors when actually in error. I'd add a second caveat, which is that most of the benefit from admitting an error is in the lessons learnt by retracing steps and finding where they went wrong. Each error has a specific cause - a doubt not investigated, a piece of evidence given too much or too little weight, or a bias triggered. I try to make myself stronger by identifying those causes, concretely envisioning what I should have done differently, and thinking of the reference classes where the same mistake might happen in the future.

The wording actually given in this quote avoids the problems discussed by Will_Newsome and jimrandomh: admitting error clears the score, resets it to zero. If you were wrong, this wipes out your negative score, for a net win; if you were right, it wipes out your positive score, setting you back.

(Unless you weren't in error. Once you start awarding yourself internal karma for admitting that you were wrong, it becomes much easier to do so even when you weren't actually wrong. Of course, this is sidestepped with empiricism.)