Selfish mining paper, for reference: https://arxiv.org/pdf/1311.0243v1.pdf
Some stats from my experiment with selfish mining over a period 753 blocks:
- min time received compared to header timestamp: -26 sec (so the header timestamp was 26 seconds in the future. lol)
- max time received compared to header timestamp: 166 sec
- median: 0
- mean: 3.34661
- stddev: 11.896
No apparently selfish blocks were captured during this time period.
The selfish mining threshold is slightly under one standard deviation. Let me know if you think this value should be larger, just keep in mind that it's only used for breaking ties.