Tie the tail to -9 volts.
Tie the top of the 2 collector resistors to +9v.
Notice this requires a total of 18 volts. Call the midpoint "ground"
Now tie both bases to GND thru 1Kohm resistors, to supply the base current.
Set the tail current to 2 milliamps (thus approximately 1ma per transistor).
Put 4,700 ohm resistors in each collector, with the top ends to +9v supply.
With 4.7 volts across each of the two collector resistors, we see each transistor has about 4.3 volts collector-base. And the common emitter node is about -0.6 volts. Approximately.
You might have fun varying the tail current, in factors of 2:1 or 4:1 or 8:1 or 10:1, and using log-linear plot paper to examine the logarithmic behavior of that emitter voltage as the current widely changes.
Apply 1mvpp to either base. Unless that 1mv is precisely centered around ground, you may want to insert a series 100uf between the sin generator and the base. Use a much faster frequency than 1 Hertz.
What gain? for small signals (and 1mv qualifies as small signal), the delta_Iout / delta_Vbase for each of the 2 transistors is 1/26 ohms if operating at 1ma.
When you wiggle 1 base at 1mvpp, the shared emitters wiggle at 1/2 of that Vin or 0.5mv; and the other base should be almost still (should be << 1 millivolt); the changing base current causes this undriven base to move.
Your gain, looking at either output collector, will be 4700 / (26 + 26) ,
approximately. The predictability depends on how close to exact abrupt junction the transistor emitter-base junction is manufactured.
For additional learning, perform a SPICE model of this, and examine the THREE input voltages (the driven input, the common_emitters, and the undriven input). Also examine how closely the two collector voltages are exact opposites (or are not exactly opposite); above 10,000,000 Hertz input, you will be significant imbalances; as you understand the causes of imbalances, circuit design houses will be interested in hiring you.