When you bid on a steem bid bot, you expect, that the round length is as declared.
144 Minutes long.
But is this always the case?
I looked into the voting patterns of all bid bots known to me and took their data of the last 30 days.
I searched for all voting transactions of every bot and took the time between those transactions.
Everything below 8000 seconds I did not count, most of the time deltas is about 1 minute, the time between votes, after a biding window.
from this data ( >8000 seconds) I took the median and the average mean and put it in this list.
This list is sorted by median, ascending.
This is very revealing.
It seems, that a lot of bots wait, until they are profitable and execute the voting.
My own bot, 
I consider this not good, but ok.
I have no idea, why the mean is over 150 for my bot, as it runs in a 3 minutes cycle.
Sometimes, the RPC fails, and the bot has currently 0 profit, nothing to run my own RPC on.
But everything over 150 on median and mean I would consider cheating not really nice for the user, but in most cases necessary;).
I don't expect any owners of those bots to answer me.
But you, dear reader can make your own guesses.
And resteem this, or if you want, I can give you more details/codes/data, to continue this topic on your own and post your own analysis.
We wait to 100% every round, like all the other bots. If our votes take up more than 2% that means it will take longer to recharge.
When Steem was 0.86, we would get 30-60 bids per window and most were higher than the minimum bid. Bidding windows were extremely consistent and almost perfectly 10 a day.
Now that Steem has climbed 500% we receive a lot more bids as it is still profitable. A majority of the bids are the minimum amount which gives a better return due to rounding errors. This increases how long it takes to bid which isn't really a big deal until it took over an hour to do bids when it used to take 2-5 minutes.
Due to the fact, the number of bids increased by about 700%, the percentage of each vote went down dramatically and causes more voting power to be used (due to the rounding issues mentioned) per $1 upvoted.
That is the main cause of the windows taking longer, I have increased minimum bids to 2 SBD and now 2.5 SBD to decrease the unique bids per window. It still isn't perfect and we still drain more than what a 100% full vote would drain.
We doing the best we can with the tools available. As I said before when we raised our bid to 2 SBD, mathematically, our minimum bid should be 5 SBD, but I don't want to make it unavailable to small users.
There is nothing malicious going on, at least in our case, it's just the inefficiencies showing themselves as things scale.
Thanks, that explains a lot for me!
:)
