The age group that includes 60 yr olds
Have the most experience to learn good from bad and this group is most for trump and they turn out to vote. At 71%
The least experienced group that includes the 18 yr old has the least experience to learn good from bad and they vote most for liberals and they turn out the least of all groups. 42%
So this shows only polls that weighs correctly likely voters are legit
Like rasmussen
A likely voter poll is one in which the pollster tries to estimate the turnout. It can be based on history but it can be based on other factors as well. Since Trump was elected, polls have underestimated Democrat support.
Final Results -- -- 53.3 44.9 Democrats +8.4
RCP Average 10/13 - 11/3 -- 49.7 42.4 Democrats +7.3
NBC News/Wall St. Jrnl 11/1 - 11/3 774 LV 50 43 Democrats +7
CNN 11/1 - 11/3 1151 LV 55 42 Democrats +13
ABC News/Wash Post 10/29 - 11/1 737 LV 52 44 Democrats +8
Rasmussen Reports 10/28 - 11/1 2500 LV 45 46 Republicans +1
IBD/TIPP 10/25 - 11/3 798 LV 50 41 Democrats +9
Economist/YouGov 10/28 - 10/30 1296 RV 47 42 Democrats +5
NPR/PBS/Marist 10/28 - 10/29 509 LV 52 43 Democrats +9
Harvard-Harris 10/26 - 10/28 1835 RV 47 38 Democrats +9
Reuters/Ipsos 10/24 - 10/30 1383 LV 49 42 Democrats +7
USA Today/Suffolk 10/18 - 10/22 1000 LV 51 43 Democrats +8
FOX News 10/13 - 10/16 841 LV 49 42 Democrats +7
These are the final polls for the 2018 midterms. There was very little difference between registered and likely voters. You also see most of the pollsters got it very close. The 2 exceptions were CNN which had a larger margin for Democrats than actual turnout. The second was Rasmussen which got the party ahead wrong and missed by 9.4 points. Clearly their estimate of who was going to vote was completely wrong.