Two ways you can do this: 1- Scrap 1 minute data and get tick data direct from the exchange. With good data you will see every quote (ticks are irrelevant). Then you can see the spread 100% and do your stats. 2- Do a statistical run and compare bar close with next bar open. If it's 50% of the time higher or lower by one tick, that means the spread is 1 tick 100% of the time. If the difference is 2 ticks 10% of the time, that means the spread is 2 ticks 20% of the time. This is stochastic sampling and you need a lot of 1-minute data. The more data you have obviously, the lower margin of error (confidence interval). This assumes you are talking about the usual 1-minute trade data (not midpoint or something).

from your topic what does "Estimating the Spread From 1 Minute Data" even mean? the bid/offer spread? the avg spread between 2 contracts?

Maybe you mean, look at the ticks left over after removing zero-ticks (which would be the great majority of ticks, for instruments that trade actively)? Hard to be sure, since you didn't clarify. In any case, even that wouldn't be quite right, because some fluctuations of >1 tick would be because the market went 10-11, 11-12, 12-13 (or whatever) without a trade, so the difference between successive prints would be >1 tick even though the bid-ask was 1 tick at all times.

Apparently there are ways, I found this stuff in the meantime... http://www.nber.org/public_html/confer/2009/mms09/Corwin_Schultz.pdf This paper also mentions the 'roll' covariance estimator, which is supposedly quite well known.