I don't go along with the views Sobers was the best all-rounder, I would class him as a batsman who was a very good bowler.
I tried a rating system to compare all-rounders, only those with 1000+ runs, 100+ wickets, a Test hundred and at least one 5wi to their name - it's tough! Reason it is tough is because it is unlikely someone is a pure all-rounder, who does exactly as much with bat as ball. Some, like Sobers and Hadlee, will be so strong in one discipline over the other that you have a hard time doing it so that those who say average 33 with bat and 24 with ball get more credit than someone who is runs or wicket heavy - that is without excluding the likes of Sobers and Hadlee on the premise they weren't true all-rounders.
While we class Sobers, Hadlee, Flintoff and Kallis (to name a few) as all-rounders, do they really stack up as such? Sobers, Kallis and Flintoff have never taken a 10wm haul, Hadlee has scored hundreds but two hundreds to NINE 10wm hauls is hardly equal contribution. And others have scored hundreds without being all-rounders, sadly I had no choice but to include Jason Gillespie as he meets the four main requirements.
So here's what I have currently as the top 10, although I admit I haven't checked for potential newbs.
Top 10 all-rounders?
1. Keith Miller (AUS, 55 Tests) : 2958 runs @ 36.98 & 170 wkts @ 22.98
100 x7, 50 x13 (HS 147). 5wi x7, 10wm x1 (SR 52.81, BB 7/60)
2. Ian Botham (ENG, 102 Tests) : 5200 runs @ 33.55 & 383 wkts @ 28.40
100 x14, 50 x22 (HS 208). 5wi x27, 10wm x4 (SR 55.61, BB 8/34)
3. Imran Khan (PAK, 88 Tests) : 3807 runs @ 37.69 & 362 wkts @ 22.81
100 x6, 50 x18 (HS 136). 5wi x23, 10wm x6 (SR 51.48, BB 8/58)
4. Chris Cairns (NZE, 62 Tests) : 3320 runs @ 33.54 & 218 wkts @ 29.40
100 x5, 50 x22 (HS 158). 5wi x13, 10wm x1 (SR 53.66, BB 7/27)
5. Tony Greig (ENG, 58 Tests) : 3599 runs @ 40.44 & 141 wkts @ 32.21
100 x8, 50 x20 (HS 148). 5wi x6, 10wm x2 (SR 66.23, BB 8/86)
6. Montague Noble (AUS, 42 Tests) : 1997 runs @ 30.26 & 121 wkts @ 25.00
100 x1, 50 x16 (HS 133). 5wi x9, 10wm x2 (SR 60.17, BB 7/17)
7. Wilfred Rhodes (ENG, 58 Tests) : 2325 runs @ 30.19 & 127 wkts @ 26.97
100 x2, 50 x11 (HS 179). 5wi x6, 10wm x1 (SR 65.96, BB 8/68)
8. Irfan Pathan (IND, 29 Tests) : 1105 runs @ 31.57 & 100 wkts @ 32.26
100 x1, 50 x6 (HS 102). 5wi x7, 10wm x2 (SR 58.84, BB 7/59)
9. Richard Hadless (NZE, 86 Tests) : 3124 runs @ 27.17 & 431 wkts @ 22.30
100 x2, 50 x15 (HS 151no). 5wi x36, 10wm x9 (SR 48.18, BB 9/52)
10. Kapil Dev (IND, 131 Tests) : 5248 runs @ 31.05 & 434 wkts @ 29.65
100 x8, 50 x27 (HS 163). 5wi x23, 10wm x2 (SR 63.92, BB 9/83)
As I recall it compares runs per innings, percentages of 100s, 50s, 5wi and 10wm per match and volume of runs or wickets are not in any way factored in ie if a bowler takes 5 wickets per match then that is treated the same if one has 200 wickets and one has 300 wickets, they would be distinguished between by a different factor.
If I can crack a way to reward all-roundness over bat or bowl heavy stats then I'll be delighted, perhaps standard deviations from average or something like that. If it is possible to add an extra factor for differentiating exactly the same stats where one has taken twice the number of wickets and scored twice the number of runs, but at the same rates, percentage 10wms etc then even better.
You can immediately see Hadlee is top heavy with bowling, two hundreds yet 36 5wis and his bowling figures are way better than his batting figures. Kallis comes in at 12, only five 5wis and while at a good average, that does him no favours. Flintoff came in at 19 for similar reasons and with weaker batting.
Sobers came in at 13, a batting average of 57.78 excessive and FIFTY-SIX 50s and 100s combined to just SIX 5wis. He also took only 2.5 wickets per match, although that is better than Greig (Greig too two 10fers though)
If you upped the entry requirements to make it the "elite" and made a 10wm a minimum requirement then Kallis, Sobers, Lindwall, Flintoff and a few others wouldn't make it in. 100s and 10wms are probably not quite equals, so you'd want to exclude anyone with less than say one 100 for every 20 innings played which would eliminate Noble, Rhodes, Pathan and Hadlee.
DISCLAIMER : no effort was made to try and get the result desired, other than to try and achieve the effect as mentioned above (more all-roundness) I neither agree nor disagree with the results, they are just what came out.