Just to update with some more data.
Code:
| calculator | year | 1 | 1b | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---------------|------|-----|-----|-----|-----|-----|-----|-----|-----|-----|
| Plus42 1.1.9 | 2022 | -29 | -32 | -28 | -18 | -∞ | -30 | -31 | -26 | -35 |
| NSTK TVM V5 | 2023 | -29 | -30 | -28 | -18 | -10 | -30 | -31 | -19 | -30 |
| TI-83 Plus | 1999 | -10 | -12 | -8 | -7 | -∞ | -11 | -11 | -19 | -18 |
| RPN83P v0.9.0 | 2023 | -9 | -12 | -9 | -7 | -∞ | ERR | ERR | -11 | -12 |
| HP-12cp | 2003 | -8 | -10 | -6 | -6 | -12 | ERR | ERR | -21 | -11 |
| HP-17BII | 1990 | -8 | -10 | -6 | -6 | -12 | ERR | ERR | -11 | -11 |
| HP-20b⁴ | 2008 | -8 | -10 | -6 | -6 | -12 | ERR | ERR | -11 | -11 |
| HP-30b⁴ | 2010 | -8 | -10 | -6 | -6 | -12 | ERR | ERR | -11 | -11 |
| Casio FC-200 | 1992 | -6 | -8 | NA | -4 | -13 | ERR | ERR | -19 | -9 |
| HP-92³ | 1977 | -6 | -8 | -4 | -4 | -∞ | ERR | ERR | -9 | -9 |
| HP-12c⁴ | 1981 | -6 | -8 | -4 | -4 | -10 | ERR | ERR | -9 | -11 |
| HP-37E¹ | 1978 | -6 | -8 | -4 | -4 | -9 | ERR | ERR | -9 | -9 |
| HP-38E¹ | 1978 | -6 | -8 | -4 | -4 | -9 | ERR | ERR | -9 | -9 |
| HP-27¹ | 1976 | -6 | -9 | -4 | -3 | -9 | ERR | ERR | ERR | ERR |
| HP-22¹ | 1975 | -6 | -8 | -4 | -3 | -9 | ERR | ERR | ERR | ERR |
| HP-70¹ | 1974 | -5 | -6 | -4 | 5 | -9 | ERR | ERR | ERR | ERR |
| HP-80² | 1973 | -6 | -4 | -5 | 5 | ERR | ERR | ERR | ERR | ERR |
I've used a combination of contributions and emulated data. It is interesting to see the progression of the accuracy, with more capability of the calculators to solve particular types of problems as the algorithm is improved. It was also interesting to see the huge UI improvements over time, with the HP-92 essentially landing on the peak UI style, where you can chain calculations easily, and it adopts the modern sign conventions.