There are much better motivations for log-odds and the logistic function than just ensuring the predicted values are bounded probabilities. If you want to do b

There are much better motivations for log-odds and the logistic function than just ensuring the predicted values are bounded probabilities. If you want to do bayesian updating of your model upon seeing new data, and you want to express this as a linear function, a sum of log-odds values is the way to do it. The best way to understand this is to learn about "naive bayes" first. If you start with naive bayes, but replace the values that are based on observed frequencies with values that you can just optimize for your objective, you get logistic regression.